|
(11) | EP 4 351 145 A2 |
| (12) | EUROPEAN PATENT APPLICATION |
|
|
|
|
|||||||||||||||||||||||
|
||||||||||||||||||||||||
| (54) | VIDEO ENCODING DEVICE, VIDEO ENCODING METHOD, VIDEO ENCODING PROGRAM, VIDEO DECODING DEVICE, VIDEO DECODING METHOD, AND VIDEO DECODING PROGRAM |
| (57) There is provided a technique that includes a triangle merging candidate list constructor
structured to construct a triangle merging candidate list including spatial merging
candidates, a first triangle merging candidate selector structured to select, from
the triangle merging candidate list, a first triangle merging candidate that is uni-prediction,
and a second triangle merging candidate selector structured to select, from the triangle
merging candidate list, a second triangle merging candidate that is uni-prediction,
in which in a region where motion compensation by weighted averaging by the first
triangle merging candidate and the second triangle merging candidate is performed,
uni-prediction motion information of one of the first triangle merging candidate or
the second triangle merging candidate is saved.
|
BACKGROUND
[0001] The present invention relates to picture coding and decoding techniques in which a picture is split into blocks and prediction is performed.
[0002] In coding and decoding of a picture, a target picture is split into blocks, each of which being a group of a predetermined number of samples, and processing is performed in units of blocks. Splitting a picture into appropriate blocks with appropriate settings of intra prediction and inter prediction enables improvement of coding efficiency.
[0003] Coding/decoding of a moving picture uses inter prediction that performs prediction from a coded/decoded picture, thereby improving coding efficiency. Patent Literature 1 describes a technique of applying an affine transform at the time of inter prediction. Moving pictures often involve object deformation such as enlargement/reduction or rotation, and thus, application of the technique in Patent Document 1 enables efficient coding.
SUMMARY
[0005] Unfortunately, however, the technique of Patent Document 1 involves picture transform, leading to a problem of heavy processing load. The present invention has been made in view of the above problem, and provides a low load and efficient coding technique.
[0006] In one aspect of the present invention to solve the above problem, there is provided a technique that includes: a triangle merging candidate list constructor structured to construct a triangle merging candidate list including spatial merging candidates; a first triangle merging candidate selector structured to select, from the triangle merging candidate list, a first triangle merging candidate that is uni-prediction; and a second triangle merging candidate selector structured to select, from the triangle merging candidate list, a second triangle merging candidate that is uni-prediction, in which in a region where motion compensation by weighted averaging by the first triangle merging candidate and the second triangle merging candidate is performed, uni-prediction motion information of one of the first triangle merging candidate or the second triangle merging candidate is saved.
[0007] According to the present invention, it is possible to achieve highly efficient and low load picture coding/decoding process.
BRIEF DESCRIPTION OF THE DRAWINGS
[0008]
Fig. 1 is a block diagram of a picture coding device according to an embodiment of the present invention.
Fig. 2 is a block diagram of a picture decoding device according to an embodiment of the present invention.
Fig. 3 is a flowchart illustrating operation of splitting a tree block.
Fig. 4 is a diagram illustrating a state of splitting an input picture into tree blocks.
Fig. 5 is a diagram illustrating z-scan.
Fig. 6A is a diagram illustrating a split shape of a block.
Fig. 6B is a diagram illustrating a split shape of a block.
Fig. 6C is a diagram illustrating a split shape of a block.
Fig. 6D is a diagram illustrating a split shape of a block.
Fig. 6E is a diagram illustrating a split shape of a block.
Fig. 7 is a flowchart illustrating operation of splitting a block into four.
Fig. 8 is a flowchart illustrating operation of splitting a block into two or three.
Fig. 9 is syntax for expressing the shape of block split.
Fig. 10A is a diagram illustrating intra prediction.
Fig. 10B is a diagram illustrating intra prediction.
Fig. 11 is a diagram illustrating reference blocks for inter prediction.
Fig. 12A is syntax for expressing a coding block prediction mode.
Fig. 12B is syntax for expressing the coding block prediction mode.
Fig. 13 is a diagram illustrating a correspondence between syntax elements and modes related to inter prediction.
Fig. 14 is a diagram illustrating affine motion compensation at two control points.
Fig. 15 is a diagram illustrating affine motion compensation at three control points.
Fig. 16 is a block diagram of a detailed configuration of an inter prediction unit 102 in Fig. 1.
Fig. 17 is a block diagram of a detailed configuration of a normal motion vector predictor mode derivation unit 301 in Fig. 16.
Fig. 18 is a block diagram of a detailed configuration of a normal merge mode derivation unit 302 in Fig. 16.
Fig. 19 is a flowchart illustrating a normal motion vector predictor mode derivation process of the normal motion vector predictor mode derivation unit 301 in Fig. 16.
Fig. 20 is a flowchart illustrating a processing procedure of the normal motion vector predictor mode derivation process.
Fig. 21 is a flowchart illustrating a processing procedure of a normal merge mode derivation process.
Fig. 22 is a block diagram of a detailed configuration of an inter prediction unit 203 in Fig. 2.
Fig. 23 is a block diagram of a detailed configuration of a normal motion vector predictor mode derivation unit 401 in Fig. 22.
Fig. 24 is a block diagram of a detailed configuration of a normal merge mode derivation unit 402 in Fig. 22.
Fig. 25 is a flowchart illustrating a normal motion vector predictor mode derivation process of a normal motion vector predictor mode derivation unit 401 in Fig. 22.
Fig. 26 is a diagram illustrating a history-based motion vector predictor candidate list initialization/update processing procedure.
Fig. 27 is a flowchart of an identical element confirmation processing procedure in the history-based motion vector predictor candidate list initialization/update processing procedure.
Fig. 28 is a flowchart of an element shift processing procedure in the history-based motion vector predictor candidate list initialization/update processing procedure.
Fig. 29 is a flowchart illustrating a history-based motion vector predictor candidate derivation processing procedure.
Fig. 30 is a flowchart illustrating a history-based merging candidate derivation processing procedure.
Fig. 31A is a diagram illustrating an example of a history-based motion vector predictor candidate list update process.
Fig. 31B is a diagram illustrating an example of a history-based motion vector predictor candidate list update process.
Fig. 31C is a diagram illustrating an example of a history-based motion vector predictor candidate list update process.
Fig. 32 is a diagram illustrating motion compensation prediction in a case where LO-prediction is performed and a reference picture (RefL0Pic) of L0 is at a time before a target picture (CurPic).
Fig. 33 is a diagram illustrating motion compensation prediction in a case where LO-prediction is performed and a reference picture of LO-prediction is at a time after the target picture.
Fig. 34 is a diagram illustrating a prediction direction of motion compensation prediction in bi-prediction in which an LO-prediction reference picture is at a time before the target picture and an L1-prediction reference picture is at a time after the target picture.
Fig. 35 is a diagram illustrating a prediction direction of motion compensation prediction in bi-prediction in which an LO-prediction reference picture and an L1-prediction reference picture are at a time before the target picture.
Fig. 36 is a diagram illustrating a prediction direction of motion compensation prediction in bi-prediction in which an LO-prediction reference picture and an L1-prediction reference picture are at a time after the target picture.
Fig. 37 is a diagram illustrating an example of a hardware configuration of a coding-decoding device according to an embodiment of the present invention.
Fig. 38A is a diagram illustrating prediction of a triangle merge mode.
Fig. 38B is a diagram illustrating prediction of the triangle merge mode.
Fig. 39 is a flowchart illustrating an average merging candidate derivation processing procedure.
Fig. 40 is a flowchart illustrating triangle merging candidate derivation.
Fig. 41 is a flowchart illustrating derivation of uni-prediction motion information of a merge triangle partition 0 according to the present embodiment.
Fig. 42 is a flowchart illustrating derivation of uni-prediction motion information of a merge triangle partition 1 in the embodiment of the present invention.
Fig. 43A is a diagram illustrating weighting in the triangle merge mode.
Fig. 43B is a diagram illustrating weighting in the triangle merge mode.
Fig. 44A is a diagram illustrating partitions in the triangle merge mode.
Fig. 44B is a diagram illustrating partitions in the triangle merge mode.
Fig. 44C is a diagram illustrating partitions in the triangle merge mode.
Fig. 44D is a diagram illustrating partitions in the triangle merge mode.
Fig. 44E is a diagram illustrating partitions in the triangle merge mode.
Fig. 44F is a diagram illustrating partitions in the triangle merge mode.
Fig. 45A is a diagram illustrating stored information in the triangle merge mode.
Fig. 45B is a diagram illustrating stored information in the triangle merge mode.
Fig. 46A is a diagram illustrating stored information in the triangle merge mode.
Fig. 46B is a diagram illustrating stored information in the triangle merge mode.
Fig. 47A is a diagram illustrating stored information in the triangle merge mode.
Fig. 47B is a diagram illustrating stored information in the triangle merge mode.
Fig. 48A is a diagram illustrating stored information in the triangle merge mode.
Fig. 48B is a diagram illustrating stored information in the triangle merge mode.
DETAILED DESCRIPTION
Tree block
[0010] In the embodiment, a coding/decoding process target picture (processing target picture) is equally split into a predetermined size. This unit is defined as a tree block. While Fig. 4 sets the size of the tree block to 128 × 128 samples, the size of the tree block is not limited to this and may be set to any size. The target tree block (corresponding to a coding target in the coding process and a decoding target in the decoding process) is switched in raster scan order, that is, in order from left to right and from top to bottom. The interior of each tree block can be further recursively split. A coding/decoding block as a result of recursive split of the tree block is defined as a coding block. A tree block and a coding block are collectively defined as a block. Execution of appropriate block split enables efficient coding. The size of the tree block may be a fixed value determined in advance by the coding device and the decoding device, or it is possible to adopt a configuration in which the size of the tree block determined by the coding device is transmitted to the decoding device. Here, the maximum size of the tree block is 128 × 128 samples, and the minimum size of the tree block is 16 × 16 samples. The maximum size of the coding block is 64 × 64 samples, and the minimum size of the coding block is 4 × 4 samples.
Prediction mode
[0011] Switching is performed between intra prediction (MODE_INTRA) of performing prediction from a processed picture signal of a target picture and inter prediction (MODE_INTER) of performing prediction from a picture signal of a processed picture in units of target coding blocks.
[0012] The processed picture is used, in the coding process, for a picture obtained by decoding a coded signal, a picture signal, a tree block, a block, a coding block, or the like. The processed picture is used, in the decoding process, for a decoded picture, picture signal, a tree block, a block, a coding block, or the like.
[0013] A mode of identifying the intra prediction (MODE_INTRA) and the inter prediction (MODE_INTER) is defined as a prediction mode (PredMode). The prediction mode (PredMode) has intra prediction (MODE_INTRA) or inter prediction (MODE_INTER) as a value.
Inter prediction
[0014] In inter prediction in which prediction is performed from a picture signal of a processed picture, it is possible to use a plurality of processed pictures as reference pictures. In order to manage a plurality of reference pictures, two types of reference lists L0 (reference list 0) and L1 (reference list 1) are defined. A reference picture is specified using a reference index in each of the lists. In the P slice, LO-prediction (Pred_L0) is usable. In the B slice, LO-prediction (Pred_L0), L1-prediction (Pred_L1), and bi-prediction (Pred_BI) is usable. LO-prediction (Pred_L0) is inter prediction that refers to a reference picture managed by L0, while L1-prediction (Pred_L1) is inter prediction that refers to a reference picture managed by L1. Bi-prediction (Pred_BI) is inter prediction in which both LO-prediction and L1-prediction are performed and one reference picture managed in each of L0 and L1 is referred to. Information specifying LO-prediction, L1-prediction, and bi-prediction is defined as an inter prediction mode. In the following processing, it is assumed that processing will be performed for each of L0 and L1 for constants and variables including a suffix LX in an output.
Motion vector predictor mode
[0015] The motion vector predictor mode is a mode of transmitting an index for specifying a motion vector predictor, a motion vector difference, an inter prediction mode, and a reference index, and determining inter prediction information of a target block. The motion vector predictor is derived from a motion vector predictor candidate derived from a processed block in the neighbor of the target block or a block belonging to the processed picture and located at the same position as or in the neighborhood (vicinity) of the target block, and from an index to specify the motion vector predictor.
Merge mode
[0016] The merge mode is a mode that derives inter prediction information of the target block from inter prediction information of a processed block in the neighbor of the target block, or a block belonging to a processed picture and located at the same position as the target block or in the neighborhood (vicinity) of the target block, without transmit a motion vector difference or a reference index.
[0017] A processed block in the neighbor of the target block and inter prediction information of the processed block are defined as spatial merging candidates. Blocks belonging to the processed picture and located at the same position as the target block or in the neighborhood (vicinity) of the target block, and inter prediction information derived from the inter prediction information of the block are defined as temporal merging candidates. Each of merging candidates is registered in a merging candidate list. A merging candidate to be used for prediction of a target block is specified by a merge index.
Neighboring block
[0018] Fig. 11 is a diagram illustrating reference blocks to be referred to for deriving inter prediction information in the motion vector predictor mode and the merge mode. A0, A1, A2, B0, B1, B2, and B3 are processed blocks in the neighbor of the target block. T0 is a block belonging to the processed picture and located at the same position as the target block or in the neighborhood (vicinity) of the target block, in the target picture.
[0019] A1 and A2 are blocks located on the left side of the target coding block and in the neighbor of the target coding block. B1 and B3 are blocks located above the target coding block and in the neighbor of the target coding block. A0, B0, and B2 are blocks respectively located at the lower left, the upper right, and the upper left of the target coding block.
[0020] Details of how neighboring blocks are handled in the motion vector predictor mode and the merge mode will be described below.
Affine motion compensation
[0021] The affine motion compensation first splits a coding block into subblocks of a predetermined unit and then individually determines a motion vector for each of the split subblocks to perform motion compensation. The motion vector of each of subblocks is derived on the basis of one or more control points derived from the inter prediction information of a processed block in the neighbor of the target block, or a block belonging to the processed picture and located at the same position as or in the neighborhood (vicinity) of the target block. While the present embodiment sets the size of the subblock to 4 × 4 samples, the size of the subblock is not limited to this, and a motion vector may be derived in units of samples.
[0022] Fig. 14 illustrates an example of affine motion compensation in a case where there are two control points. In this case, each of the two control points has two parameters, that is, a horizontal component and a vertical component. Accordingly, the affine transform having two control points is referred to as four-parameter affine transform. CP1 and CP2 in Fig. 14 are control points.
[0023] Fig. 15 illustrates an example of affine motion compensation in a case where there are three control points. In this case, each of the three control points has two parameters, that is, a horizontal component and a vertical component. Accordingly, the affine transform having three control points is referred to as six-parameter affine transform. CP1, CP2, and CP3 in Fig. 15 are control points.
[0024] The affine motion compensation is usable in any of the motion vector predictor mode and the merge mode. A mode of applying the affine motion compensation in the motion vector predictor mode is defined as a subblock motion vector predictor mode. A mode of applying the affine motion compensation in the merge mode is defined as a subblock merge mode.
Syntax of coding block
[0025] The syntax for expressing the prediction mode of the coding block will be described with reference to Figs. 12A, 12B, and 13. The pred_mode_flag in Fig. 12A is a flag indicating whether the mode is inter prediction. Setting of pred_mode_flag 0 indicates inter prediction while setting of pred_mode_flag 1 indicates intra prediction. Information of intra prediction intra_pred_mode is transmitted in the case of intra prediction, while merge_flag is transmitted in the case of inter prediction. merge_flag is a flag indicating whether the mode to use is the merge mode or the motion vector predictor mode. In the case of the motion vector predictor mode (merge_flag = 0), a flag inter_affine_flag indicating whether to apply the subblock motion vector predictor mode is transmitted. In the case of applying the subblock motion vector predictor mode (inter_affine_flag = 1), cu_affine_type_flag is transmitted. cu_affine_type_flag is a flag for determining the number of control points in the subblock motion vector predictor mode.
[0026] In contrast, in the case of the merge mode (merge_flag = 1), the merge_subblock_flag of Fig. 12B is transmitted. merge_subblock_flag is a flag indicating whether to apply the subblock merge mode. In the case of the subblock merge mode (merge_subblock_flag = 1), a merge index merge_subblock_idx is transmitted. Conversely, in a case where the mode is not the subblock merge mode (merge_subblock_flag = 0), a flag merge_triangle_flag indicating whether to apply the triangle merge mode is transmitted. In the case of applying the triangle merge mode (merge_triangle_flag = 1), merge triangle indexes merge_triangle_idx0 and merge_triangle_idx1 are transmitted for each of the block splitting directions merge_triangle_split_dir, and for each of the two split partitions. In the case of not applying the triangle merge mode, (merge_triangle_flag = 0), a merge index merge_idx is transmitted.
[0027] Fig. 13 illustrates the value of each of syntax elements and the corresponding prediction mode. merge_flag = 0 and inter_affine_flag = 0 correspond to the normal motion vector predictor mode (Inter Pred Mode). merge_flag = 0 and inter_affine_flag = 1 correspond to a subblock motion vector predictor mode (Inter Affine Mode). merge_flag = 1, merge_subblock_flag = 0, and merge_trianlge_flag = 0 correspond to a normal merge mode (Merge Mode). merge_flag = 1, merge_subblock_flag = 0, and merge_trianlge_flag = 1 correspond to a triangle merge mode (Triangle Merge Mode). merge_flag = 1, merge_subblock_flag = 1 correspond to a subblock merge mode (Affine Merge Mode).
POC
[0028] A Picture Order Count (POC) is a variable associated with the picture to be coded, and is set to a value that increments by one in accordance with picture output order. The POC value makes it possible to discriminate whether the pictures are the same, discriminate inter-picture sequential relationship in the output order, or derive the distance between the pictures. For example, it is possible to determine that two pictures having a same POC value are identical pictures. In a case where the POCs of the two pictures have different values, the picture with the smaller POC value can be determined to be the picture that is output earlier. The difference between the POCs of the two pictures indicates the distance between the pictures in the time axis direction.
First Embodiment
[0029] A picture coding device 100 and a picture decoding device 200 according to a first embodiment of the present invention will be described.
[0030] Fig. 1 is a block diagram of the picture coding device 100 according to the first embodiment. The picture coding device 100 according to an embodiment includes a block split unit 101, an inter prediction unit 102, an intra prediction unit 103, decoded picture memory 104, a prediction method determiner 105, a residual generation unit 106, an orthogonal transformer/quantizer 107, a bit strings coding unit 108, an inverse quantizer/inverse orthogonal transformer 109, a decoded picture signal superimposer 110, and coding information storage memory 111.
[0031] The block split unit 101 recursively splits an input picture to construct a coding block. The block split unit 101 includes: a quad split unit that splits a split target block in both the horizontal direction and the vertical direction; and a binary-ternary split unit that splits a split target block in either the horizontal direction or the vertical direction. The block split unit 101 sets the constructed coding block as a target coding block, and supplies a picture signal of the target coding block to the inter prediction unit 102, the intra prediction unit 103, and the residual generation unit 106. Further, the block split unit 101 supplies information indicating the determined recursive split structure to the bit strings coding unit 108. Detailed operation of the block split unit 101 will be described below.
[0032] The inter prediction unit 102 performs inter prediction of the target coding block. The inter prediction unit 102 derives a plurality of inter prediction information candidates from the inter prediction information stored in the coding information storage memory 111 and the decoded picture signal stored in the decoded picture memory 104, selects a suitable inter prediction mode from the plurality of derived candidates, and supplies the selected inter prediction mode and a predicted picture signal corresponding to the selected inter prediction mode to the prediction method determiner 105. Detailed configuration and operation of the inter prediction unit 102 will be described below.
[0033] The intra prediction unit 103 performs intra prediction on the target coding block. The intra prediction unit 103 refers to the decoded picture signal stored in the decoded picture memory 104 as a reference sample, and performs intra prediction based on coding information such as an intra prediction mode stored in the coding information storage memory 111 and thereby generates a predicted picture signal. In the intra prediction, the intra prediction unit 103 selects a suitable intra prediction mode from a plurality of intra prediction modes, and supplies the selected intra prediction mode and the selected predicted picture signal corresponding to the selected intra prediction mode to the prediction method determiner 105.
[0034] Figs. 10A and 10B illustrate examples of intra prediction. Fig. 10A illustrates a correspondence between the prediction direction of intra prediction and the intra prediction mode number. For example, an intra prediction mode 50 copies reference samples in the vertical direction and thereby constructs an intra prediction picture. Intra prediction mode 1 is a DC mode in which all sample values of a target block are set to an average value of reference samples. Intra prediction mode 0 is a Planar mode in which a two-dimensional intra prediction picture is created from reference samples in the vertical and horizontal directions. Fig. 10B is an example of constructing an intra prediction picture in the case of an intra prediction mode 40. The intra prediction unit 103 copies, for each of samples of the target block, the value of the reference sample in the direction indicated by the intra prediction mode. In a case where the reference sample in the intra prediction mode is not at an integer position, the intra prediction unit 103 determines a reference sample value by interpolation from reference sample values at neighboring integer positions.
[0035] The decoded picture memory 104 stores the decoded picture constructed by the decoded picture signal superimposer 110. The decoded picture memory 104 supplies the stored decoded picture to the inter prediction unit 102 and the intra prediction unit 103.
[0036] The prediction method determiner 105 evaluates each of the intra prediction and the inter prediction using the coding information, the code amount of the residual, the distortion amount between the predicted picture signal and the target picture signal, or the like, and thereby determines an optimal prediction mode. In the case of intra prediction, the prediction method determiner 105 supplies intra prediction information such as an intra prediction mode to the bit strings coding unit 108 as coding information. In the case of the merge mode of the inter prediction, the prediction method determiner 105 supplies inter prediction information such as a merge index and information (subblock merge flag) indicating whether the mode is the subblock merge mode to the bit strings coding unit 108 as coding information. In the case of the motion vector predictor mode of the inter prediction, the prediction method determiner 105 supplies inter prediction information such as the inter prediction mode, the motion vector predictor index, the reference index of L0 or L1, the motion vector difference, or information indicating whether the mode is a subblock motion vector predictor mode (subblock motion vector predictor flag) to the bit strings coding unit 108 as coding information. The prediction method determiner 105 further supplies the determined coding information to the coding information storage memory 111. The prediction method determiner 105 supplies the predicted picture signal to the residual generation unit 106 and the decoded picture signal superimposer 110.
[0037] The residual generation unit 106 constructs a residual by subtracting the predicted picture signal from the target picture signal, and supplies the constructed residual to the orthogonal transformer/quantizer 107.
[0038] The orthogonal transformer/quantizer 107 performs orthogonal transform and quantization on the residual according to the quantization parameter and thereby constructs an orthogonally transformed and quantized residual, and then supplies the constructed residual to the bit strings coding unit 108 and the inverse quantizer/inverse orthogonal transformer 109.
[0039] The bit strings coding unit 108 codes, in addition to the sequences, pictures, slices, and information in units of coding blocks, the bit strings coding unit 108 encodes coding information corresponding to the prediction method determined by the prediction method determiner 105 for each of coding blocks. Specifically, the bit strings coding unit 108 encodes a prediction mode PredMode for each of coding blocks. In a case where the prediction mode is inter prediction (MODE_INTER), the bit strings coding unit 108 encodes coding information (inter prediction information) such as a flag to determine whether the mode is the merge mode, a subblock merge flag, a merge index in the case of merge mode, an inter prediction mode in the case of non-merge modes, a motion vector predictor index, information related to motion vector differences, and a subblock motion vector predictor flag, on the bases of a prescribed syntax (bit string syntax rules) and thereby constructs a first bit string. In a case where the prediction mode is intra prediction (MODE_INTRA), coding information (intra prediction information) such as the intra prediction mode is coded according to a prescribed syntax (bit string syntax rules) to construct a first bit string. In addition, the bit strings coding unit 108 performs entropy coding on the orthogonally transformed and quantized residual on the basis of a prescribed syntax and thereby constructs a second bit string. The bit strings coding unit 108 multiplexes the first bit string and the second bit string on the basis of a prescribed syntax, and outputs the bitstream.
[0040] The inverse quantizer/inverse orthogonal transformer 109 performs inverse quantization and inverse orthogonal transform on the orthogonally transformed/quantized residual supplied from the orthogonal transformer/quantizer 107 and thereby calculates the residual, and then supplies the calculated residual to the decoded picture signal superimposer 110.
[0041] The decoded picture signal superimposer 110 superimposes the predicted picture signal according to the determination of the prediction method determiner 105 with the residual that has undergone the inverse quantization/inverse orthogonal transform by the inverse quantizer/inverse orthogonal transformer 109, thereby constructs a decoded picture, and stores the constructed decoded picture in the decoded picture memory 104. The decoded picture signal superimposer 110 may perform filtering processing of reducing distortion such as block distortion due to coding on the decoded picture, and may thereafter store the decoded picture in the decoded picture memory 104.
[0042] The coding information storage memory 111 stores coding information such as a prediction mode (inter prediction or intra prediction) determined by the prediction method determiner 105. In the case of inter prediction, the coding information stored in the coding information storage memory 111 includes inter prediction information such as the determined motion vector, reference indexes of the reference lists L0 and L1, and a history-based motion vector predictor candidate list. In the case of the inter prediction merge mode, the coding information stored in the coding information storage memory 111 includes, in addition to the above-described information, a merge index and inter prediction information including information indicating whether the mode is a subblock merge mode (a subblock merge flag). In the case of the motion vector predictor mode of the inter prediction, the coding information stored in the coding information storage memory 111 includes, in addition to the above information, inter prediction information such as an inter prediction mode, a motion vector predictor index, a motion vector difference, and information indicating whether the mode is a subblock motion vector predictor mode (subblock motion vector predictor flag). In the case of intra prediction, the coding information stored in the coding information storage memory 111 includes intra prediction information such as the determined intra prediction mode.
[0043] Fig. 2 is a block diagram illustrating a configuration of a picture decoding device according to an embodiment of the present invention corresponding to the picture coding device of Fig. 1. The picture decoding device according to the embodiment includes a bit strings decoding unit 201, a block split unit 202, an inter prediction unit 203, an intra prediction unit 204, coding information storage memory 205, an inverse quantizer/inverse orthogonal transformer 206, and a decoded picture signal superimposer 207, and decoded picture memory 208.
[0044] Since the decoding process of the picture decoding device in Fig. 2 corresponds to the decoding process provided inside the picture coding device in Fig. 1. Accordingly, each of configurations of the coding information storage memory 205, the inverse quantizer/inverse orthogonal transformer 206, the decoded picture signal superimposer 207, and the decoded picture memory 208 in Fig. 2 respectively has a function corresponding to each of the configurations of the coding information storage memory 111, the inverse quantizer/inverse orthogonal transformer 109, the decoded picture signal superimposer 110, and the decoded picture memory 104 of the picture coding device in Fig. 1.
[0045] The bitstream supplied to the bit strings decoding unit 201 is separated on the basis of a prescribed syntax rule. The bit strings decoding unit 201 decodes the separated first bit string, and thereby obtains sequence, a picture, a slice, information in units of coding blocks, and coding information in units of coding blocks. Specifically, the bit strings decoding unit 201 decodes a prediction mode PredMode that discriminates whether the prediction is inter prediction (MODE_INTER) or intra prediction (MODE_INTRA) in units of coding block. In a case where the prediction mode is the inter prediction (MODE_INTER), the bit strings decoding unit 201 decodes coding information (inter prediction information) related to the flag that discriminates whether the mode is the merge mode, the merge index in the case of the merge mode, the subblock merge flag, and the inter prediction mode in the case of the motion vector predictor mode, the motion vector predictor index, motion vector difference, the subblock motion vector predictor flag or the like according to a prescribed syntax, and then, supplies the coding information (inter prediction information) to the coding information storage memory 205 via the inter prediction unit 203 and the block split unit 202. In a case where the prediction mode is intra prediction (MODE_INTRA), the bit strings decoding unit 201 decodes coding information (intra prediction information) such as the intra prediction mode according to a prescribed syntax, and then supplies the decoded coding information (intra prediction information) to the coding information storage memory 205 via the inter prediction unit 203 or the intra prediction unit 204, and via the block split unit 202. The bit strings decoding unit 201 decodes the separated second bit string and calculates an orthogonally transformed/quantized residual, and then, supplies the orthogonally transformed/quantized residual to the inverse quantizer/inverse orthogonal transformer 206.
[0046] When the prediction mode PredMode of the target coding block is the inter prediction (MODE_INTER) and the motion vector predictor mode, the inter prediction unit 203 uses the coding information of the already decoded picture signal stored in the coding information storage memory 205 to derive a plurality of motion vector predictor candidates. The inter prediction unit 203 then registers the plurality of derived motion vector predictor candidates to a motion vector predictor candidate list described below. The inter prediction unit 203 selects a motion vector predictor corresponding to the motion vector predictor index to be decoded and supplied by the bit strings decoding unit 201 from among the plurality of motion vector predictor candidates registered in the motion vector predictor candidate list. The inter prediction unit 203 then calculates a motion vector on the basis of the motion vector difference decoded by the bit strings decoding unit 201 and the selected motion vector predictor, and stores the calculated motion vector in the coding information storage memory 205 together with other coding information. Here, the coding information of the coding block to be supplied and stored includes the prediction mode PredMode, flags predFlagL0[xP][yP] and predFlagL1[xP][yP] indicating whether to use LO-prediction and L1-prediction, reference indexes refIdxL0[xP][yP] and refIdxL1[xP][yP] of L0 and L1; and motion vectors mvL0[xP][yP] and mvL1[xP][yP] of L0 and L1, or the like. Here, xP and yP are indexes indicating the position of the upper left sample of the coding block within the picture. In a case where the prediction mode PredMode is inter prediction (MODE_INTER) and the inter prediction mode is LO-prediction (Pred_L0), the flag predFlagLO indicating whether to use LO-prediction is set to 1 and the flag predFlagL1 indicating whether to use L1-prediction is set to 0. In a case where the inter prediction mode is L1-prediction (Pred_L1), a flag predFlagLO indicating whether to use LO-prediction is set to 0 and a flag predFlagL1 indicating whether to use L1-prediction is set to 1. In a case where the inter prediction mode is bi-prediction (Pred_BI), both the flag predFlagLO indicating whether to use LO-prediction and the flag predFlagL1 indicating whether to use L1-prediction are set to 1. Furthermore, when the prediction mode PredMode of the target coding block is in the inter prediction (MODE_INTER) and the merge mode, a merging candidate is derived. Using the coding information of the already-decoded coding block stored in the coding information storage memory 205, a plurality of merging candidates is derived and registered in a merging candidate list described below. Subsequently, a merging candidate corresponding to the merge index that is decoded by the bit strings decoding unit 201 and supplied is selected from among the plurality of merging candidates registered in the merging candidate list, and then, inter prediction information such as flags predFlagL0[xP][yP] and predFlagL1[xP][yP] indicating whether to use the LO-prediction and L1-prediction of the selected merging candidate, reference indexes refIdxL0[xP][yP] and refIdxL1[xP][yP] of L0 and L1, and motion vectors mvL0[xP][yP] and mvL1[xP][yP] of L0 and L1 are to be stored in the coding information storage memory 205. Here, xP and yP are indexes indicating the position of the upper left sample of the coding block within the picture. Detailed configuration and operation of the inter prediction unit 203 will be described below.
[0047] The intra prediction unit 204 performs intra prediction when the prediction mode PredMode of the target coding block is intra prediction (MODE_INTRA). The coding information decoded by the bit strings decoding unit 201 includes an intra prediction mode. The intra prediction unit 204 generates a predicted picture signal by intra prediction from the decoded picture signal stored in the decoded picture memory 208 in accordance with the intra prediction mode included in the coding information decoded by the bit strings decoding unit 201. The intra prediction unit 204 then supplies the generated predicted picture signal to the decoded picture signal superimposer 207. The intra prediction unit 204 corresponds to the intra prediction unit 103 of the picture coding device 100, and thus performs the processing similar to the processing of the intra prediction unit 103.
[0048] The inverse quantizer/inverse orthogonal transformer 206 performs inverse orthogonal transform/inverse quantization on the orthogonal transformed/quantized residual decoded in the bit strings decoding unit 201, and thereby obtains inversely orthogonally transformed/inversely quantized residual.
[0049] The decoded picture signal superimposer 207 superimposes a predicted picture signal inter-predicted by the inter prediction unit 203 or a prediction picture signal intra-predicted by the intra prediction unit 204 with the residual that has been inversely orthogonally transformed/inversely quantized residual by the inverse quantizer/inverse orthogonal transformer 206, thereby decoding the decoded picture signal. The decoded picture signal superimposer 207 then stores the decoded picture signal that has been decoded, in the decoded picture memory 208. When storing the decoded picture in the decoded picture memory 208, the decoded picture signal superimposer 207 may perform filtering processing on the decoded picture to reduce block distortion or the like due to coding, and may thereafter store the decoded picture in the decoded picture memory 208.
[0050] Next, operation of the block split unit 101 in the picture coding device 100 will be described. Fig. 3 is a flowchart illustrating operation of splitting a picture into tree blocks and further splitting each of the tree blocks. First, an input picture is split into tree blocks of a predetermined size (step S1001). Each of the tree blocks is scanned in a predetermined order, that is, in a raster scan order (step S1002), and a target tree block is internally split (step S1003).
[0051] Fig. 7 is a flowchart illustrating detailed operation of the split process in step S1003. First, it is determined whether to split the target block into four (step S1101) .
[0052] In a case where it is determined that the target block is to be split into four, the target block will be split into four (step S1102). Each of blocks obtained by splitting the target block is scanned in the Z-scan order, that is, in the order of upper left, upper right, lower left, and lower right (step S1103). Fig. 5 illustrates an example of the Z-scan order, and 601 in Fig. 6A illustrates an example in which the target block is split into four. Numbers 0 to 3 of 601 in Fig. 6A indicate the order of processing. Subsequently, the split process of Fig. 7 is recursively executed for each of blocks split in step S1101 (step S1104).
[0053] In a case where it is determined that the target block is not to be split into four, the target block will be split into two or three, namely, binary-ternary split (step S1105).
[0054] Fig. 8 is a flowchart illustrating detailed operation of the binary-ternary split process in step S1105. First, it is determined whether binary-ternary split is going to be performed on the target block, that is, whether any of binary split or ternary split is to be performed (step S1201).
[0055] In a case where it is not determined that binary-ternary split is to be performed on the target block, that is, in a case where it is determined not to split the target block, the split is finished (step S1211). That is, further recursive split process is not to be performed on the block that has been split by the recursive split process.
[0056] In a case where it is determined that binary-ternary split is going to be performed on the target block, it is further determined whether to split the target block into two (step S1202).
[0057] In a case where it is determined that the target block is to be split into two, it is further determined whether to split the target block in upper-lower (vertical) direction (step S1203), and then based on the result, the target block will be binary split in upper-lower (vertical) direction (step S1204), or the target block will be binary split in left-right (horizontal) direction (step S1205). As a result of step S1204, the target block is binary split in upper-lower direction (vertical direction) as illustrated in 602 of Fig.6B. As a result of step S1205, the target block is binary split in right-left (horizontal direction) as illustrated in 604 of Fig.6D.
[0058] In step S1202, in a case where it is not determined that the target block is to be split into two, that is, in a case where it is determined that the target block is to be split into three, it is further determined whether to split the target block into three as upper, middle, lower portions (vertical direction) (step S1206). Based on the result, the target block is split into three as either upper, middle and lower portions (vertical direction) (step S1207), or left, middle, and right portions (horizontal direction) (step S1208). As a result of step S1207, the target block is split into three as upper, middle, and lower portions (vertical direction) as illustrated in 603 of Fig.6C. As a result of step S1208, the target block is split into three as left, middle, and right (horizontal direction) as illustrated in 605 of Fig. 6E.
[0059] After execution of one of steps S1204, S1205, S1207, or S1208, each of blocks obtained by splitting the target block is scanned in order from left to right and from top to bottom (step S1209). The numbers 0 to 2 of 602 to 605 in Figs. 6B to 6E indicate the order of processing. For each of split blocks, the binary-ternary split process in Fig. 8 is recursively executed (step S1210).
[0060] In the recursive block split described here, the propriety of split may be limited on the basis of the number of splits, the size of the target block, or the like. The information that restricts the propriety of split may be realized in a configuration in which information is not transmitted by making a preliminary agreement between the coding device and the decoding device, or in a configuration in which the coding device determines information for restricting the propriety of split and record the information to bit strings, thereby transmitting the information to the decoding device.
[0061] When a certain block is split, a block before split is referred to as a parent block, and each of blocks after split is referred to as a child block.
[0062] Next, operation of the block split unit 202 in the picture decoding device 200 will be described. The block split unit 202 splits a tree block using a processing procedure similar to case of the block split unit 101 of the picture coding device 100. Note that there is a difference that although the block split unit 101 of the picture coding device 100 determines an optimal block split shape by applying an optimization method such as estimation of an optimal shape by picture recognition or distortion rate optimization, the block split unit 202 of the picture decoding device 200 determines the block split shape by decoding block split information recorded in the bit string.
[0063] Fig. 9 illustrates syntax (bit string syntax rules) related to block split according to the first embodiment. coding_quadtree() represents the syntax for quad split process of the block. multi_type_tree() represents the syntax for the process of splitting the block into two or three. qt_split is a flag indicating whether to split a block into four. In the case of splitting the block into four, the setting would be qt_split = 1. In the case of not splitting the block into four, the setting would be qt_split = 0. In the case of splitting the block into four (qt_split = 1), quad split process will be performed recursively on each of blocks split into four (coding_quadtree (0), coding_quadtree (1), coding_quadtree (2), and coding_quadtree (3), in which arguments 0 to 3 correspond to numbers of 601 in Fig. 6A). In a case where the quad split is not to be performed (qt_split = 0), the subsequent split is determined according to multi_type_tree(). mtt_split is a flag indicating whether to perform further split. In the case where further splitting is to be performed (mtt_split = 1), transmission of mtt_split_vertical which is a flag indicating whether to perform split in vertical or horizontal direction and mtt_split_binary which is a flag that determines whether to split the block into two or three will be performed. mtt_split_vertical = 1 indicates split in the vertical direction, and mtt_split_vertical = 0 indicates split in the horizontal direction. mtt_split_binary = 1 indicates that the block is binary split, and mtt_split_binary = 0 indicates that the block is ternary split. In a case where the block is to be binary split (mtt_split_binary = 1), the split process is performed recursively on each of the two split blocks (multi_type_tree (0) and multi_type_tree (1) in which arguments 0 to 1 correspond to numbers in 602 or 604 of Figs. 6B to 6D). In the case where the block is to be ternary split (mtt_split_binary = 0), the split process is performed recursively on each of the three split blocks (multi_type_tree (0), multi_type_tree (1), and multi_type_tree (2), in which 0 to 2 correspond to numbers in 603 of Fig. 6B or 605 of Fig. 6E). Recursively calling multi_type_tree until mtt_split = 0 will achieve hierarchical block split.
Inter prediction
[0064] The inter prediction method according to an embodiment is implemented in the inter prediction unit 102 of the picture coding device in Fig. 1 and the inter prediction unit 203 of the picture decoding device in Fig. 2.
[0065] An inter prediction method according to an embodiment will be described with reference to the drawings. The inter prediction method is implemented in any of the coding and decoding processes in units of coding blocks.
Inter prediction unit 102 on the coding side
[0066] Fig. 16 is a diagram illustrating a detailed configuration of the inter prediction unit 102 of the picture coding device in Fig. 1. The normal motion vector predictor mode derivation unit 301 derives a plurality of normal motion vector predictor candidates, selects a motion vector predictor, and calculates a motion vector difference between the selected motion vector predictor and the detected motion vector. The detected inter prediction mode, reference index, motion vector, and calculated motion vector difference will be inter prediction information of the normal motion vector predictor mode. This inter prediction information is supplied to the inter prediction mode determiner 305. Detailed configuration and processing of the normal motion vector predictor mode derivation unit 301 will be described below.
[0067] The normal merge mode derivation unit 302 derives a plurality of normal merging candidates, selects a normal merging candidate, and obtains inter prediction information of the normal merge mode. This inter prediction information is supplied to the inter prediction mode determiner 305. Detailed configuration and processing of the normal merge mode derivation unit 302 will be described below.
[0068] The subblock motion vector predictor mode derivation unit 303 derives a plurality of subblock motion vector predictor candidates, selects a subblock motion vector predictor, and calculates a motion vector difference between the selected subblock motion vector predictor and the detected motion vector. The detected inter prediction mode, reference index, motion vector, and calculated motion vector difference will be inter prediction information of the subblock motion vector predictor mode. This inter prediction information is supplied to the inter prediction mode determiner 305.
[0069] The subblock merge mode derivation unit 304 derives a plurality of subblock merging candidates, selects a subblock merging candidate, and obtains inter prediction information of the subblock merge mode. This inter prediction information is supplied to the inter prediction mode determiner 305.
[0070] In the inter prediction mode determiner 305 determines inter prediction information on the basis of the inter prediction information supplied from the normal motion vector predictor mode derivation unit 301, the normal merge mode derivation unit 302, the subblock motion vector predictor mode derivation unit 303, and the subblock merge mode derivation unit 304. Inter prediction information according to the determination result is supplied from the inter prediction mode determiner 305 to a motion compensation prediction unit 306.
[0071] The motion compensation prediction unit 306 performs inter prediction on the reference picture signal stored in the decoded picture memory 104 on the basis of the determined inter prediction information. Detailed configuration and processing of the motion compensation prediction unit 306 will be described below.
Inter prediction unit 203 on decoding side
[0072] Fig. 22 is a diagram illustrating a detailed configuration of the inter prediction unit 203 of the picture decoding device in Fig. 2.
[0073] The normal motion vector predictor mode derivation unit 401 derives a plurality of normal motion vector predictor candidates, selects a motion vector predictor, calculates an added value obtained by adding the selected motion vector predictor and the decoded motion vector difference, and sets this added value as a motion vector. The decoded inter prediction mode, reference index, motion vector will be inter prediction information of the normal motion vector predictor mode. This inter prediction information is supplied to the motion compensation prediction unit 406 via the switch 408. Detailed configuration and processing of the normal motion vector predictor mode derivation unit 401 will be described below.
[0074] The normal merge mode derivation unit 402 derives a plurality of normal merging candidates, selects a normal merging candidate, and obtains inter prediction information of the normal merge mode. This inter prediction information is supplied to the motion compensation prediction unit 406 via the switch 408. Detailed configuration and processing of the normal merge mode derivation unit 402 will be described below.
[0075] A subblock motion vector predictor mode derivation unit 403 derives a plurality of subblock motion vector predictor candidates, selects a subblock motion vector predictor, and calculates an added value obtained by adding the selected subblock motion vector predictor and the decoded motion vector difference, and sets this added value as a motion vector. The decoded inter prediction mode, reference index, and motion vector will be the inter prediction information of the subblock motion vector predictor mode. This inter prediction information is supplied to the motion compensation prediction unit 406 via the switch 408.
[0076] A subblock merge mode derivation unit 404 derives a plurality of subblock merging candidates, selects a subblock merging candidate, and obtains inter prediction information of the subblock merge mode. This inter prediction information is supplied to the motion compensation prediction unit 406 via the switch 408.
[0077] The motion compensation prediction unit 406 performs inter prediction on the reference picture signal stored in the decoded picture memory 208 on the basis of the determined inter prediction information. Detailed configuration and processing of the motion compensation prediction unit 406 are similar to the motion compensation prediction unit 306 on the coding side.
Normal motion vector predictor mode derivation unit (Normal AMVP)
[0078] The normal motion vector predictor mode derivation unit 301 of Fig. 17 includes a spatial motion vector predictor candidate derivation unit 321, a temporal motion vector predictor candidate derivation unit 322, a history-based motion vector predictor candidate derivation unit 323, a motion vector predictor candidate replenisher 325, a normal motion vector detector 326, a motion vector predictor candidate selector 327, and a motion vector subtractor 328.
[0079] The normal motion vector predictor mode derivation unit 401 in Fig. 23 includes a spatial motion vector predictor candidate derivation unit 421, a temporal motion vector predictor candidate derivation unit 422, a history-based motion vector predictor candidate derivation unit 423, a motion vector predictor candidate replenisher 425, a motion vector predictor candidate selector 426, and a motion vector adder 427.
[0080] Processing procedures of the normal motion vector predictor mode derivation unit 301 on the coding side and the normal motion vector predictor mode derivation unit 401 on the decoding side will be described with reference to the flowcharts in Figs. 19 and 25, respectively. Fig. 19 is a flowchart illustrating a normal motion vector predictor mode derivation processing procedure performed by the normal motion vector mode derivation unit 301 on the coding side. Fig. 25 is a flowchart illustrating a normal motion vector predictor mode derivation processing procedure performed by the normal motion vector mode derivation unit 401 on the decoding side.
Normal motion vector predictor mode derivation unit (Normal AMVP): coding side
[0081] The normal motion vector predictor mode derivation processing procedure on the coding side will be described with reference to Fig. 19. In the description of the processing procedure in Fig. 19, the word "normal" illustrated in Fig. 19 will be omitted in some cases.
[0082] First, the normal motion vector detector 326 detects a normal motion vector for each of inter prediction modes and reference indexes (step S100 in Fig. 19).
[0083] Subsequently, a motion vector difference of a motion vector used in inter prediction in the normal motion vector predictor mode is calculated for each of L0 and L1 (steps S101 to S106 in Fig. 19) in the spatial motion vector predictor candidate derivation unit 321, the temporal motion vector predictor candidate derivation unit 322, the history-based motion vector predictor candidate derivation unit 323, the motion vector predictor candidate replenisher 325, the motion vector predictor candidate selector 327, and the motion vector subtractor 328. Specifically, in a case where the prediction mode PredMode of the target block is inter prediction (MODE_INTER) and the inter prediction mode is L0-prediction (Pred_L0), the motion vector predictor candidate list mvpListLO of L0 is calculated. Subsequently, the motion vector predictor mvpLO is selected, and then, a motion vector difference mvdL0 of the motion vector mvLO of L0 is calculated. In a case where the inter prediction mode of the target block is L1-prediction (Pred_L1), a motion vector predictor candidate list mvpListL1 of L1 is calculated. Subsequently, a motion vector predictor mvpL1 is selected, and then a motion vector difference mvdL1 of a motion vector mvL1 of L1 is calculated. In a case where the inter prediction mode of the target block is bi-prediction (Pred_BI), LO-prediction and L1-prediction are both performed. A motion vector predictor candidate list mvpListLO of L0 is calculated and a motion vector predictor mvpLO of L0 is selected, and then a motion vector difference mvdL0 of the motion vector mvLO of L0 is calculated. Along with this calculation, a motion vector predictor candidate list mvpListL1 of L1 is calculated and a motion vector predictor mvpL1 of L1 is calculated, and then, a motion vector difference mvdL1 of a motion vector mvL1 of L1 is calculated.
[0084] The motion vector difference calculation process is performed for each of L0 and L1, in which the calculation process is a common process in both L0 and L1. Accordingly, L0 and L1 will be denoted as LX as a common procedure. In the process of calculating the motion vector difference of L0, X of LX is set to 0, while in the process of calculating the motion vector difference of L1, X of LX is set to 1. Additionally, in a case where information on the other list is referred to instead of one LX during the calculation process of the motion vector difference of the one LX, the other list will be represented as LY.
[0085] In a case where a motion vector mvLX of LX is used (step S102 in Fig. 19: YES), motion vector predictor candidates of LX are calculated, thereby constructing a motion vector predictor candidate list mvpListLX of LX (step S103 in Fig. 19). In the normal motion vector predictor mode derivation unit 301, the spatial motion vector predictor candidate derivation unit 321, the temporal motion vector predictor candidate derivation unit 322, the history-based motion vector predictor candidate derivation unit 323, and the motion vector predictor candidate replenisher 325 derive a plurality of motion vector predictor candidates and thereby constructs the motion vector predictor candidate list mvpListLX. The detailed processing procedure of step S103 in Fig. 19 will be described below using the flowchart in Fig. 20.
[0086] Subsequently, the motion vector predictor candidate selector 327 selects a motion vector predictor mvpLX of LX from the motion vector predictor candidate list mvpListLX of LX (step S104 in Fig. 19). Here, one element (the i-th element counted from 0) in the motion vector predictor candidate list mvpListLX is represented as mvpListLX[i]. The motion vector difference, which is a difference between the motion vector mvLX and each of the motion vector predictor candidates mvpListLX[i] stored in the motion vector predictor candidate list mvpListLX, is each calculated. A code amount at the time of coding these motion vector differences is calculated for each of elements (motion vector predictor candidates) of the motion vector predictor candidate list mvpListLX. Subsequently, the motion vector predictor candidate mvpListLX[i] that minimizes the code amount for each of motion vector predictor candidates among the individual elements registered in the motion vector predictor candidate list mvpListLX is selected as the motion vector predictor mvpLX, and its index i is obtained. In a case where there is a plurality of the motion vector predictor candidates having the minimum generated code amount in the motion vector predictor candidate list mvpListLX, the motion vector predictor candidate mvpListLX[i] having the index i in the motion vector predictor candidate list mvpListLX represented by a small number is selected as the optimal motion vector predictor mvpLX, and its index i is obtained.
[0087] Subsequently, the motion vector subtractor 328 subtracts the selected motion vector predictor mvpLX of LX from the motion vector mvLX of LX and calculates a motion vector difference mvdLX of LX as in: mvdLX=mvLX-mvpLX (step S105 in Fig. 19).
Normal motion vector predictor mode derivation unit (normal AMVP): decoding side
[0088] Next, a normal motion vector predictor mode processing procedure on the decoding side will be described with reference to Fig. 25. On the decoding side, the spatial motion vector predictor candidate derivation unit 421, the temporal motion vector predictor candidate derivation unit 422, the history-based motion vector predictor candidate derivation unit 423, and the motion vector predictor candidate replenisher 425 individually calculate motion vectors used in the inter prediction of the normal motion vector predictor mode for each of L0 and L1 (steps S201 to S206 in Fig. 25). Specifically, in a case where the prediction mode PredMode of the target block is inter prediction (MODE_INTER) and the inter prediction mode of the target block is LO-prediction (Pred_L0), the motion vector predictor candidate list mvpListLO of L0 is calculated. Subsequently, the motion vector predictor mvpLO is selected, and then, the motion vector mvLO of L0 is calculated. In a case where the inter prediction mode of the target block is L1-prediction (Pred_L1), the L1 motion vector predictor candidate list mvpListL1 is calculated. Subsequently, the motion vector predictor mvpL1 is selected, and the L1 motion vector mvL1 is calculated. In a case where the inter prediction mode of the target block is bi-prediction (Pred_BI), LO-prediction and L1-prediction are both performed. A motion vector predictor candidate list mvpListLO of L0 is calculated and a motion vector predictor mvpLO of L0 is selected, and then the motion vector mvLO of L0 is calculated. Along with this calculation, a motion vector predictor candidate list mvpListL1 of L1 is calculated and a motion vector predictor mvpL1 of L1 is calculated, and then, the motion vector mvL1 of L1 is calculated.
[0089] Similarly to the coding side, the decoding side performs the motion vector calculation processing for each of L0 and L1, in which the processing is a common process in both L0 and L1. Accordingly, L0 and L1 will be denoted as LX as a common procedure. LX represents an inter prediction mode used for inter prediction of a target coding block. X is 0 in the process of calculating the motion vector of L0, and X is 1 in the process of calculating the motion vector of L1. Additionally, in a case where information on the other reference list is referred to instead of the same reference list as the LX to be calculated during the calculation process of the motion vector of the LX, the other reference list will be represented as LY.
[0090] In a case where the motion vector mvLX of LX is used (step S202 in Fig. 25: YES), motion vector predictor candidates of LX are calculated to construct a motion vector predictor candidate list mvpListLX of LX (step S203 in Fig. 25). In the normal motion vector predictor mode derivation unit 401, the spatial motion vector predictor candidate derivation unit 421, the temporal motion vector predictor candidate derivation unit 422, the history-based motion vector predictor candidate derivation unit 423, and the motion vector predictor candidate replenisher 425 calculate a plurality of motion vector predictor candidates and thereby constructs the motion vector predictor candidate list mvpListLX. Detailed processing procedure of step S203 in Fig. 25 will be described below using the flowchart in Fig. 20.
[0091] Subsequently, the motion vector predictor candidate selector 426 extracts a motion vector predictor candidate mvpListLX[mvpIdxLX] corresponding to the motion vector predictor index mvpIdxLX decoded and supplied by the bit strings decoding unit 201 from the motion vector predictor candidate list mvpListLX, as the selected motion vector predictor mvpLX (step S204 in Fig. 25).
[0092] Subsequently, the motion vector adder 427 adds the motion vector difference mvdLX of LX decoded and supplied by the bit strings decoding unit 201 and the motion vector predictor mvpLX of LX and calculates a motion vector mvLX of LX as in: mvLX=mvpLX+mvdLX (step S205 in Fig. 25).
Normal motion vector predictor mode derivation unit (normal AMVP): motion vector prediction method
[0093] Fig. 20 is a flowchart illustrating a processing procedure of the normal motion vector predictor mode derivation process having a function common to the normal motion vector predictor mode derivation unit 301 of the picture coding device and the normal motion vector predictor mode derivation unit 401 of the picture decoding device according to the embodiment of the present invention.
[0094] Each of the normal motion vector predictor mode derivation unit 301 and the normal motion vector predictor mode derivation unit 401 includes a motion vector predictor candidate list mvpListLX. The motion vector predictor candidate list mvpListLX has a list structure, and includes a storage region that stores, as elements, a motion vector predictor index indicating a location in the motion vector predictor candidate list and a motion vector predictor candidate corresponding to the index. The number of the motion vector predictor index starts from 0, and motion vector predictor candidates are to be stored in the storage region of the motion vector predictor candidate list mvpListLX. In the present embodiment, it is assumed that the motion vector predictor candidate list mvpListLX can register at least two motion vector predictor candidates (as inter prediction information). Further, a variable numCurrMvpCand indicating the number of motion vector predictor candidates registered in the motion vector predictor candidate list mvpListLX is set to 0.
[0095] Each of the spatial motion vector predictor candidate derivation units 321 and 421 derives a motion vector predictor candidate from blocks in the neighbor of the left side. This process derives a motion vector predictor mvLXA with reference to inter prediction information of the block in the neighbor of the left side (AC or A1 in Fig. 11), namely, a flag indicating whether a motion vector predictor candidate is usable, a motion vector, a reference index, or the like, and adds the derived mvLXA to the motion vector predictor candidate list mvpListLX (step S301 in Fig. 20). Note that X is 0 in LO-prediction and X is 1 in L1-prediction (similar applies hereinafter). Subsequently, the spatial motion vector predictor candidate derivation units 321 and 421 derive the motion vector predictor candidates from an upper neighboring block. This process derives a motion vector predictor mvLXB with reference to inter prediction information of the upper neighboring block (B0, B1 or B2 in Fig. 11), namely, a flag indicating whether a motion vector predictor candidate is usable, a motion vector, a reference index, or the like, When the derived mvLXA and the derived mvLXB are not equal, mvLXB is added to the motion vector predictor candidate list mvpListLX (step S302 in Fig. 20). The processes in steps S301 and S302 in Fig. 20 are provided as a common process except that the positions and numbers of reference neighboring blocks are different, and a flag availableFlagLXN indicating whether a motion vector predictor candidate of a coding block is usable, and a motion vector mvLXN, a reference index refIdxN (N indicates A or B, similar applies hereinafter) will be derived in these processes.
[0096] Subsequently, each of the temporal motion vector predictor candidate derivation units 322 and 422 derives a motion vector predictor candidate from a block in a picture having a temporal difference from the target picture. This process derives a flag availableFlagLXCol indicating whether a motion vector predictor candidate of a coding block of a picture having a temporal difference is usable, and a motion vector mvLXCol, a reference index refIdxCol, and a reference list listCol, and adds mvLXCol to the motion vector predictor candidate list mvpListLX (step S303 in Fig. 20).
[0097] Note that it is assumed that the processes of the temporal motion vector predictor candidate derivation units 322 and 422 can be omitted in units of a sequence (SPS), a picture (PPS), or a slice.
[0098] Subsequently, the history-based motion vector predictor candidate derivation units 323 and 423 add the history-based motion vector predictor candidates registered in a history-based motion vector predictor candidate list HmvpCandList to the motion vector predictor candidate list mvpListLX. (Step S304 in Fig. 20). Details of the registration processing procedure in step S304 will be described below with reference to the flowchart in Fig. 29.
[0099] Subsequently, the motion vector predictor candidate replenishers 325 and 425 add a motion vector predictor candidate having a predetermined value such as (0, 0) until the motion vector predictor candidate list mvpListLX is satisfied (S305 in Fig. 20).
Normal merge mode derivation unit (normal merge)
[0100] The normal merge mode derivation unit 302 in Fig. 18 includes a spatial merging candidate derivation unit 341, a temporal merging candidate derivation unit 342, an average merging candidate derivation unit 344, a history-based merging candidate derivation unit 345, a merging candidate replenisher 346, and a merging candidate selector 347.
[0101] The normal merge mode derivation unit 402 in Fig. 24 includes a spatial merging candidate derivation unit 441, a temporal merging candidate derivation unit 442, an average merging candidate derivation unit 444, a history-based merging candidate derivation unit 445, a merging candidate replenisher 446, and a merging candidate selector 447.
[0102] Fig. 21 is a flowchart illustrating a procedure of a normal merge mode derivation process having a function common to the normal merge mode derivation unit 302 of the picture coding device and the normal merge mode derivation unit 402 of the picture decoding device according to the embodiment of the present invention.
[0103] Hereinafter, various processes will be described step by step. The following description is a case where the slice type slice_type is B slice unless otherwise specified. However, the present invention can also be applied to the case of P slice. Note that, there is only LO-prediction (Pred_L0) as the inter prediction mode, with no L1-prediction (Pred_L1) or bi-prediction (Pred_BI) in the case where the slice type slice_type is P slice. Accordingly, it is possible to omit the process related to L1 in this case.
[0104] The normal merge mode derivation unit 302 and the normal merge mode derivation unit 402 include a merging candidate list mergeCandList. The merging candidate list mergeCandList has a list structure, and includes a storage region that stores, as elements, a merge index indicating a location in the merging candidate list and a merging candidate corresponding to the index. The number of the merge index starts from 0, and the merging candidate is stored in the storage region of the merging candidate list mergeCandList. In the subsequent processing, the merging candidate of the merge index i registered in the merging candidate list mergeCandList will be represented by mergeCandList[i]. In the present embodiment, it is assumed that the merging candidate list mergeCandList can register at least six merging candidates (as inter prediction information). Furthermore, a variable numCurrMergeCand indicating the number of merging candidates registered in the merging candidate list mergeCandList is set to 0.
[0105] The spatial merging candidate derivation unit 341 and the spatial merging candidate derivation unit 441 derive a spatial merging candidate of each of blocks (B1, A1, B0, A0, B2 in Fig. 11) in the neighbor of the target block in order of B1, A1, B0, A0, and B2, from the coding information stored either in the coding information storage memory 111 of the picture coding device or in the coding information storage memory 205 of the picture decoding device, and then, registers the derived spatial merging candidates to the merging candidate list mergeCandList (step S401 in Fig. 21). Here, N indicating one of B1, A1, B0, A0, B2 or the temporal merging candidate Col will be defined. Items to be derived include a flag availableFlagN indicating whether the inter prediction information of the block N is usable as a spatial merging candidate, a reference index refIdxL0N of L0 and a reference index refIdxLlN of L1 of the spatial merging candidate N, an LO-prediction flag predFlagL0N indicating whether LO-prediction is to be performed, an L1-prediction flag predFlagL1N indicating whether L1-prediction is to be performed, a motion vector mvL0N of L0, and a motion vector mvLlN of L1. However, since the merging candidate in the present embodiment is derived without reference to the inter prediction information of the block included in the target coding block, the spatial merging candidate using the inter prediction information of the block included in the target coding block will not be derived.
[0106] Subsequently, the temporal merging candidate derivation unit 342 and the temporal merging candidate derivation unit 442 derive temporal merging candidates from pictures having a temporal difference, and register the derived temporal merging candidates in a merging candidate list mergeCandList (step S402 in Fig. 21). Items to be derived include a flag availableFlagCol indicating whether the temporal merging candidate is usable, an LO-prediction flag predFlagL0Col indicating whether the LO-prediction of the time merging candidate is to be performed, an L1-prediction flag predFlagLlCol indicating whether the L1-prediction is to be performed, and a motion vector mvL0Col of L0, and a motion vector mvLlCol of L1.
[0107] Note that it is assumed that the processes of the temporal merging candidate derivation units 342 and 442 can be omitted in units of a sequence (SPS), a picture (PPS), or a slice.
[0108] Subsequently, the history-based merging candidate derivation unit 345 and the history-based merging candidate derivation unit 445 register the history-based motion vector predictor candidates registered in the history-based motion vector predictor candidate list HmvpCandList, to the merging candidate list mergeCandList (step S403 in Fig. 21).
[0109] In a case where the number of merging candidates numCurrMergeCand registered in the merging candidate list mergeCandList is smaller than the maximum number of merging candidates MaxNumMergeCand, the history-based merging candidate is derived with the number of merging candidates numCurrMergeCand registered in the merging candidate list mergeCandList being limited to the maximum number of merging candidates MaxNumMergeCand, and then registered to the merging candidate list mergeCandList.
[0110] Subsequently, the average merging candidate derivation unit 344 and the average merging candidate derivation unit 444 derive an average merging candidate from the merging candidate list mergeCandList, and add the derived average merging candidate to the merging candidate list mergeCandList (step S404 in Fig. 21).
[0111] In a case where the number of merging candidates numCurrMergeCand registered in the merging candidate list mergeCandList is smaller than the maximum number of merging candidates MaxNumMergeCand, the average merging candidate is derived with the number of merging candidates numCurrMergeCand registered in the merging candidate list mergeCandList being limited to the maximum number of merging candidates MaxNumMergeCand, and then registered to the merging candidate list mergeCandList.
[0112] Here, the average merging candidate is a new merging candidate including a motion vector obtained by averaging the motion vectors of the first merging candidate and the second merging candidate registered in the merging candidate list mergeCandList for each of LO-prediction and L1-prediction.
[0113] Subsequently, in the merging candidate replenisher 346 and the merging candidate replenisher 446, in a case where the number of merging candidates numCurrMergeCand registered in the merging candidate list mergeCandList is smaller than the maximum number of merging candidates MaxNumMergeCand, an additional merging candidate is derived with the number of merging candidates numCurrMergeCand registered in the merging candidate list mergeCandList being limited to the maximum number of merging candidates MaxNumMergeCand, and then registered to the merging candidate list mergeCandList (step S405 in Fig. 21). In the P slice, a merging candidate having the motion vector of a value (0, 0) and the prediction mode of LO-prediction (Pred_L0) is added with the maximum number of merging candidates MaxNumMergeCand as the upper limit. In the B slice, a merging candidate having the prediction mode of bi-prediction (Pred_BI) and the motion vector of a value (0, 0) is added. The reference index at the time of addition of a merging candidate is different from the reference index that has already been added.
[0114] Subsequently, the merging candidate selector 347 and the merging candidate selector 447 select a merging candidate from among the merging candidates registered in the merging candidate list mergeCandList. The merging candidate selector 347 on the coding side calculates the code amount and the distortion amount, and thereby selects a merging candidate, and then, supplies a merge index indicating the selected merging candidate and inter prediction information of the merging candidate to the motion compensation prediction unit 306 via the inter prediction mode determiner 305. In contrast, the merging candidate selector 447 on the decoding side selects a merging candidate based on the decoded merge index, and supplies the selected merging candidate to the motion compensation prediction unit 406. Updating history-based motion vector predictor candidate list
[0115] Next, a method of initializing and updating the history-based motion vector predictor candidate list HmvpCandList provided in the coding information storage memory 111 on the coding side and the coding information storage memory 205 on the decoding side will be described in detail. Fig. 26 is a flowchart illustrating the history-based motion vector predictor candidate list initialization/update processing procedure.
[0116] In the present embodiment, the history-based motion vector predictor candidate list HmvpCandList is updated in the coding information storage memory 111 and the coding information storage memory 205. Alternatively, a history-based motion vector predictor candidate list updating unit may be provided in the inter prediction unit 102 and the inter prediction unit 203 to update the history-based motion vector predictor candidate list HmvpCandList.
[0117] Initial settings of the history-based motion vector predictor candidate list HmvpCandList are performed at the head of the slice. On the coding side, the history-based motion vector predictor candidate list HmvpCandList is updated in a case where the normal motion vector predictor mode or the normal merge mode is selected by the prediction method determiner 105. On the decoding side, the history-based motion vector predictor candidate list HmvpCandList is updated in a case where the prediction information decoded by the bit strings decoding unit 201 is the normal motion vector predictor mode or the normal merge mode.
[0118] The inter prediction information used at the time of performing the inter prediction in the normal motion vector predictor mode or the normal merge mode is to be registered in the history-based motion vector predictor candidate list HmvpCandList, as an inter prediction information candidate hMvpCand. The inter prediction information candidate hMvpCand includes the reference index refIdxL0 of L0 and the reference index refIdxL1 of L1, the LO-prediction flag predFlagL0 indicating whether L0-prediction is to be performed, the L1-prediction flag predFlagL1 indicating whether L1-prediction is to be performed, the motion vector mvL0 of L0 and the motion vector mvL1 of L1.
[0119] In a case where there is inter prediction information having the same value as the inter prediction information candidate hMvpCand among the elements (that is, inter prediction information) registered in the history-based motion vector predictor candidate list HmvpCandList provided in the coding information storage memory 111 on the coding side and the coding information storage memory 205 on the decoding side, the element will be deleted from the history-based motion vector predictor candidate list HmvpCandList. In contrast, in a case where there is no inter prediction information having the same value as the inter prediction information candidate hMvpCand, the head element of the history-based motion vector predictor candidate list HmvpCandList will be deleted, and the inter prediction information candidate hMvpCand will be added to the end of the history-based motion vector predictor candidate list HmvpCandList.
[0120] The number of elements of the history-based motion vector predictor candidate list HmvpCandList provided in the coding information storage memory 111 on the coding side and the coding information storage memory 205 on the decoding side of the present invention is set to six.
[0121] First, the history-based motion vector predictor candidate list HmvpCandList is initialized in units of slices (step S2101 in Fig. 26). All the elements of the history-based motion vector predictor candidate list HmvpCandList are emptied at the head of the slice, and the number NumHmvpCand (current number of candidates) of history-based motion vector predictor candidates registered in the history-based motion vector predictor candidate list HmvpCandList is set to 0.
[0122] Although initialization of the history-based motion vector predictor candidate list HmvpCandList is to be performed in units of slices (first coding block of a slice), the initialization may be performed in units of pictures, tiles, or tree block rows.
[0123] Subsequently, the following process of updating the history-based motion vector predictor candidate list HmvpCandList is repeatedly performed for each of coding blocks in the slice (steps S2102 to S2107 in Fig. 26).
[0124] First, initial settings are performed in units of coding blocks. A flag identicalCandExist indicating whether an identical candidate exists is set to a value of FALSE (false), a deletion target index removeIdx indicating the deletion target candidate is set to 0 (step S2103 in Fig. 26).
[0125] It is determined whether there is an inter prediction information candidate hMvpCand to be registered (step S2104 in Fig. 26). In a case where the prediction method determiner 105 on the coding side determines the normal motion vector predictor mode or the normal merge mode, or where the bit strings decoding unit 201 on the decoding side performs decoding as the normal motion vector predictor mode or the normal merge mode, the corresponding inter prediction information is set as an inter prediction information candidate hMvpCand to be registered. In a case where the prediction method determiner 105 on the coding side determines the intra prediction mode, the subblock motion vector predictor mode or the subblock merge mode, or in a case where the bit strings decoding unit 201 on the decoding side performs decoding as the intra prediction mode, the subblock motion vector predictor mode, or the subblock merge mode, update process of the history-based motion vector predictor candidate list HmvpCandList will not be performed, and there will be no inter prediction information candidate hMvpCand to be registered. In a case where there is no inter prediction information candidate hMvpCand to be registered, steps S2105 to S2106 will be skipped (step S2104 in Fig. 26: NO). In a case where there is an inter prediction information candidate hMvpCand to be registered, the process of step S2105 and later will be performed (step S2104 in Fig. 26: YES).
[0126] Subsequently, it is determined whether individual elements of the history-based motion vector predictor candidate list HmvpCandList include an element (inter prediction information) having the same value as the inter prediction information candidate hMvpCand to be registered, that is, whether the identical element exists (step S2105 in Fig. 26). Fig. 27 is a flowchart of the identical element confirmation processing procedure. In a case where the value of the number of history-based motion vector predictor candidates NumHmvpCand is 0 (step S2121: NO in Fig. 27), the history-based motion vector predictor candidate list HmvpCandList is empty, and the identical candidate does not exist. Accordingly, steps S2122 to S2125 in Fig. 27 will be skipped, finishing the identical element confirmation processing procedure. In a case where the value of the number NumHmvpCand of the history-based motion vector predictor candidates is greater than 0 (YES in step S2121 in Fig. 27), the process of step S2123 will be repeated from a history-based motion vector predictor index hMvpIdx of 0 to NumHmvpCand-1 (steps S2122 to S2125 in Fig. 27). First, comparison is made as to whether the hMvpIdx-th element HmvpCandList[hMvpIdx] counted from 0 in the history-based motion vector predictor candidate list is identical to the inter prediction information candidate hMvpCand (step S2123 in Fig. 27). In a case where they are identical (step S2123 in Fig. 27: YES), the flag identicalCandExist indicating whether the identical candidate exists is set to a value of TRUE, and the deletion target index removeIdx indicating the position of the element to be deleted is set to a current value of the history-based motion vector predictor index hMvpIdx, and the identical element confirmation processing will be finished. In a case where they are not identical (step S2123 in Fig. 27: NO), hMvpIdx is incremented by one. In a case where the history-based motion vector predictor index hMvpIdx is smaller than or equal to NumHmvpCand-1, the processing of step S2123 and later is performed.
[0127] Returning to the flowchart of Fig. 26, the process of shifting and adding elements of the history-based motion vector predictor candidate list HmvpCandList is performed (step S2106 in Fig. 26). Fig. 28 is a flowchart of the element shift/addition processing procedure of the history-based motion vector predictor candidate list HmvpCandList in step S2106 in Fig. 26. First, it is determined whether to add a new element after removing the element stored in the history-based motion vector predictor candidate list HmvpCandList, or to add a new element without removing the element. Specifically, a comparison is made as to whether the flag identicalCandExist indicating whether the identical candidate exists is TRUE, or whether NumHmvpCand is 6 (step S2141 in Fig. 28). In a case where one of the conditions that the flag identicalCandExist indicating whether the identical candidate exists is TRUE or that the number of current candidate NumHmvpCand is 6 is satisfied (step S2141: YES in Fig. 28), the element stored in the history-based motion vector predictor candidate list HmvpCandList is removed and thereafter a new element will be added. An initial value of index i is set to a value of removeIdx + 1. The element shift process of step S2143 is repeated from this initial value to NumHmvpCand. (Steps S2142 to S2144 in Fig. 28). By copying the elements of HmvpCandList[i] to HmvpCandList[i-1], the elements are shifted forward (step S2143 in Fig. 28) and i is incremented by one (steps S2142 to S2144 in Fig. 28). Subsequently, the inter prediction information candidate hMvpCand is added to the (NumHmvpCand-1)th HmvpCandList [NumHmvpCand-1] counting from 0 that corresponds to the end of the history-based motion vector predictor candidate list (step S2145 in Fig. 28), and the element shift/addition process of the history-based motion vector predictor candidate list HmvpCandList will be finished. In contrast, in a case where none of the conditions that the flag identicalCandExist indicating whether the identical candidate exists is TRUE and that NumHmvpCand is 6 are satisfied (step S2141: NO in Fig. 28), the inter prediction information candidate hMvpCand will be added to the end of the history-based motion vector predictor candidate list without removing the element stored in the history-based motion vector predictor candidate list HmvpCandList (step S2146 in Fig. 28). Here, the end of the history-based motion vector predictor candidate list is the NumHmvpCand-th HmvpCandList [NumHmvpCand] counted from 0. Moreover, NumHmvpCand is incremented by one, and the element shift and addition process of the history-based motion vector predictor candidate list HmvpCandList are finished.
[0128] Fig. 31 is a view illustrating an example of a process of updating the history-based motion vector predictor candidate list. In a case where a new element is to be added to the history-based motion vector predictor candidate list HmvpCandList in which six elements (inter prediction information) have already been registered, the history-based motion vector predictor candidate list HmvpCandList is compared with new inter prediction information in order from the head element (Fig. 31A). When the new element has the same value as the third element HMVP2 from the head of the history-based motion vector predictor candidate list HmvpCandList, the element HMVP2 is deleted from the history-based motion vector predictor candidate list HmvpCandList and the following elements HMVP3 to HMVP5 are shifted (copied) one by one forward, and a new element is added to the end of the history-based motion vector predictor candidate list HmvpCandList (Fig.31B) to complete the update of the history-based motion vector predictor candidate list HmvpCandList (Fig.31C) .
History-based motion vector predictor candidate derivation process
[0129] Next, a method of deriving a history-based motion vector predictor candidate from the history-based motion vector predictor candidate list HmvpCandList will be described in detail. This corresponds to a processing procedure of step S304 in Fig. 20 concerning common processing performed by the history-based motion vector predictor candidate derivation unit 323 of the normal motion vector predictor mode derivation unit 301 on the coding side and the history-based motion vector predictor candidate derivation unit 423 of the normal motion vector predictor mode derivation unit 401 on the decoding side. Fig. 29 is a flowchart illustrating a history-based motion vector predictor candidate derivation processing procedure.
[0130] In a case where the current number of motion vector predictor candidates numCurrMvpCand is larger than or equal to the maximum number of elements of the motion vector predictor candidate list mvpListLX (here, 2), or the number of history-based motion vector predictor candidates NumHmvpCand is 0 (step S2201: NO in Fig. 29), the process of steps S2202 to S2209 of Fig. 29 will be omitted, and the history-based motion vector predictor candidate derivation processing procedure will be finished. In a case where the number numCurrMvpCand of the current motion vector predictor candidates is smaller than 2, which is the maximum number of elements of the motion vector predictor candidate list mvpListLX, and in a case where the value of the number NumHmvpCand of the history-based motion vector predictor candidates is greater than 0 (step S2201: YES in Fig. 29), the process of steps S2202 to S2209 in Fig. 29 will be performed.
[0131] Subsequently, the process of steps S2203 to S2208 in Fig. 29 is repeated until the index i is from 1 to a smaller value out of 4 or the number of history-based motion vector predictor candidates numCheckedHMVPCand (steps S2202 to S2209 in Fig. 29). In a case where the current number of motion vector predictor candidates numCurrMvpCand is larger than or equal to 2, which is the maximum number of elements of the motion vector predictor candidate list mvpListLX (step S2203: NO in Fig. 29), the process from steps S2204 to S2209 in Fig. 29 will be omitted and the history-based motion vector predictor candidate derivation processing procedure will be finished. In a case where the current number of motion vector predictor candidates numCurrMvpCand is smaller than 2 which is the maximum number of elements in the motion vector predictor candidate list mvpListLX (step S2203 in Fig. 29: YES), the process in step S2204 and later in Fig. 29 will be performed.
[0132] Subsequently, the process in steps S2205 to S2207 is performed for cases where Y is 0 and Y is 1 (L0 and L1) (steps S2204 to S2208 in Fig. 29). In a case where the current number of motion vector predictor candidates numCurrMvpCand is larger than or equal to 2, which is the maximum number of elements of the motion vector predictor candidate list mvpListLX (step S2205: NO in Fig. 29), the process from steps S2206 to S2209 in Fig. 29 will be omitted and the history-based motion vector predictor candidate derivation processing procedure will be finished. In a case where the current number of motion vector predictor candidates numCurrMvpCand is smaller than 2 which is the maximum number of elements in the motion vector predictor candidate list mvpListLX (step S2205: YES in Fig. 29), the process in step S2206 and later in Fig. 29 will be performed.
[0133] Next, in a case where the history-based motion vector predictor candidate list HmvpCandList includes an element having the same reference index as the reference index refIdxLX of the coding/decoding target motion vector and being different from any element of the motion vector predictor list mvpListLX (step S2206: YES in Fig. 29), a motion vector of LY of the history-based motion vector predictor candidate HmvpCandList [NumHmvpCand-i] is added to the numCurrMvpCand-th element mvpListLX[numCurrMvpCand] counting from 0 in the motion vector predictor candidate list (step S2207 in Fig. 29), and the number numCurrMvpCand of the current motion vector predictor candidates is incremented by one. In a case where there is no element in the history-based motion vector predictor candidate list HmvpCandList that has the same reference index as the reference index refIdxLX of the coding/decoding target motion vector and is different from any element of the motion vector predictor list mvpListLX (step S2206: NO in Fig. 29), the additional process in step S2207 will be skipped.
[0134] The process of steps S2205 to S2207 in Fig. 29 is performed for both L0 and L1 (steps S2204 to S2208 in Fig. 29). The index i is incremented by one, and when the index i is smaller than or equal to any of smaller value of 4 or the number of history-based motion vector predictor candidates NumHmvpCand, the process of step S2203 and later will be performed again (steps S2202 to S2209 in Fig. 29).
History-based merging candidate derivation process
[0135] The following is a detailed description of a method of deriving a history-based merging candidate from the history-based merging candidate list HmvpCandList, a procedure of the process of step S404 in Fig. 21, which is a common process of the history-based merging candidate derivation unit 345 of the normal merge mode derivation unit 302 on the coding side and the history-based merging candidate derivation unit 445 of the normal merge mode derivation unit 402 on the decoding side. Fig. 30 is a flowchart illustrating a history-based merging candidate derivation processing procedure.
[0136] First, an initialization process is performed (step S2301 in Fig. 30). Each of elements from 0 to (numCurrMergeCand -1) of isPruned[i] is set to the value of FALSE, and the variable numOrigMergeCand is set to the number numCurrMergeCand of the number of elements registered in the current merging candidate list.
[0137] Subsequently, the initial value of the index hMvpIdx is set to 1, and the additional process from step S2303 to step S2310 in Fig. 30 is repeated from this initial value to NumHmvpCand (steps S2302 to S2311 in Fig. 30). When the number numCurrMergeCand of the elements registered in the current merging candidate list is not smaller than or equal to (the maximum number of merging candidates MaxNumMergeCand-1), the merging candidates have been added to all the elements in the merging candidate list. Accordingly, the history-based merging candidate derivation process will be finished (step S2303: NO in Fig. 30) In a case where the number numCurrMergeCand of the elements registered in the current merging candidate list is smaller than or equal to (the maximum number of merging candidates MaxNumMergeCand-1), the process of step S2304 and later will be performed. sameMotion is set to a value of FALSE (step S2304 in Fig. 30). Subsequently, the initial value of the index i is set to 0, and the process of steps S2306 and S2307 in Fig. 30 is performed from this initial value to numOrigMergeCand-1 (S2305 to S2308 in Fig. 30). Comparison is performed as to whether the (NumHmvpCand-hMvpIdx)-th element HmvpCandList [NumHmvpCand-hMvpIdx] counting from 0 in the history-based motion vector prediction candidate list is the same value as the i-th element mergeCandList[i] counting from 0 in the merging candidate list (step S2306 in Fig. 30).
[0138] The merging candidates is determined to have the same value in a case where all the constituent elements (inter prediction mode, reference index, motion vector) of the merging candidate have the same value. In a case where the merging candidates have the same value and isPruned[i] is set to FALSE (step S2306: YES in Fig. 30), both sameMotion and isPruned[i] will be set to TRUE (step S2307 in Fig. 30). In a case where the values are not the same (step S2306: NO in Fig. 30), the process in step S2307 will be skipped. After completion of the repetition processing from step S2305 to step S2308 in Fig. 30, comparison is made as to whether the sameMotion is FALSE (step S2309 in Fig. 30). In a case where the sameMotion is FALSE (step S2309: YES in Fig. 30), that is, the (NumHmvpCand-hMvpIdx)-th element HmvpCandList [NumHmvpCand - hMvpIdx] counting from 0 in the history-based motion vector predictor candidate list does not exist in mergeCandList, and thus, the element HmvpCandList[NumHmvpCand - hMvpIdx] that is (NumHmvpCand - hMvpIdx)th element counted from 0 of the history-based motion vector predictor candidate list is added to mergeCandList[numCurrMergeCand] that is numCurrMergeCand-th in the merging candidate list, and numCurrMergeCand is incremented by one (step S2310 in Fig. 30). The index hMvpIdx is incremented by one (step S2302 in Fig. 30), and the process of steps S2302 to S2311 in Fig. 30 is repeated.
[0139] After completion of confirmation of all the elements in the history-based motion vector predictor candidate list or completion of addition of merging candidates to all elements in the merging candidate list, the history-based merging candidate derivation process is completed.
Average merging candidate derivation process
[0140] The following is a detailed description of a method of deriving an average merging candidate, a procedure of the process of step S403 in Fig. 21, which is a common process of the average merging candidate derivation unit 344 of the normal merge mode derivation unit 302 on the coding side and the average merging candidate derivation unit 444 of the normal merge mode derivation unit 402 on the decoding side. Fig. 39 is a flowchart illustrating an average merging candidate derivation processing procedure
[0141] First, an initialization process is performed (step S1301 in Fig. 39). The variable numOrigMergeCand is set to the number of elements numCurrMergeCand registered in the current merging candidate list.
[0142] Subsequently, scanning is performed sequentially from the top of the merging candidate list to determine two pieces of motion information. Index i indicating the first motion information is set such that index i = 0, and index j indicating the second motion information is set such that index j = 1. (Steps S1302 to S1303 in Fig. 39). When the number numCurrMergeCand of the elements registered in the current merging candidate list is not smaller than or equal to (the maximum number of merging candidates MaxNumMergeCand-1), the merging candidates have been added to all the elements in the merging candidate list. Accordingly, the history-based merging candidate derivation process will be finished (step S1304 in Fig. 39). In a case where the number numCurrMergeCand of the elements registered in the current merging candidate list is smaller than or equal to (the maximum number of merging candidates MaxNumMergeCand-1), the process of step S1305 and later will be performed.
[0143] Determination is made as to whether both the i-th motion information mergeCandList[i] of the merging candidate list and j-th motion information mergeCandList[j] of the merging candidate list are invalid (step S1305 in Fig. 39). In a case where both are invalid, the process proceeds to the next element without deriving an average merging candidate of mergeCandList[i] and mergeCandList[j]. In a case where the condition that both mergeCandList[i] and mergeCandList[j] are invalid is not satisfied, the following process is repeated with X set to 0 and 1 (steps S1306 to S1314 in Fig. 39).
[0144] Determination is made as to whether the LX prediction of mergeCandList[i] is valid (step S1307 in Fig. 39). In a case where the LX prediction of mergeCandList[i] is valid, determination is made as to whether the LX prediction of mergeCandList[j] is valid (step S1308 in Fig. 39). In a case where the LX prediction of mergeCandList[j] is valid, that is, in a case where both the LX prediction of mergeCandList[i] and the LX prediction of mergeCandList[j] are valid, a motion vector of LX prediction obtained by averaging the motion vector of LX prediction of mergeCandList[i] and the motion vector of LX prediction of mergeCandList[j] will be derived, and an average merging candidate of LX prediction having a reference index of LX prediction of mergeCandList[i] will be derived, so as to be set as LX prediction of averageCand, and the LX prediction of averageCand will be validated (step S1309 in Fig. 39). In step S1308 of Fig. 39, in a case where LX prediction of mergeCandList[j] is not valid, that is, in a case where LX prediction of mergeCandList[i] is valid and LX prediction of mergeCandList[j] is invalid, a motion vector of LX prediction of mergeCandList[i] and an average merging candidate of LX prediction having a reference index will be derived, so as to be set as LX prediction of averageCand, and the LX prediction of averageCand will be validated (step S1310 in Fig. 39). In a case where the LX prediction of mergeCandList[i] is not valid in step S1307 of Fig. 39, determination is made as to whether the LX prediction of mergeCandList[j] is valid (step S1311 of Fig. 39). In a case where LX prediction of mergeCandList[j] is valid, that is, in a case where LX prediction of mergeCandList[i] is invalid and LX prediction of mergeCandList[j] is valid, a motion vector of LX prediction of mergeCandList[j] and an average merging candidate of LX prediction having a reference index will be derived, so as to be set as LX prediction of averageCand, and the LX prediction of averageCand will be validated (step S1312 in Fig. 39). In step S1311 of Fig. 39, in a case where LX prediction of mergeCandList [j] is not valid, that is, in a case where LX prediction of mergeCandList[i] and LX prediction of mergeCandList[j] are both invalid, LX prediction of averageCand will be invalidated. (step S1312 in Fig. 39).
[0145] The average merging candidate averageCand of L0-prediction, L1-prediction or BI prediction constructed as described above is added to the numCurrMergeCand-th mergeCandList[numCurrMergeCand] of the merging candidate list, and numCurrMergeCand is incremented by one (step S1315 in Fig.39). This completes the average merging candidate derivation process.
[0146] The average merging candidate is obtained by averaging in each of the horizontal component of the motion vector and the vertical component of the motion vector.
Motion compensation prediction process
[0147] The motion compensation prediction unit 306 acquires the position and size of a block that is currently subjected to prediction processing in coding. Further, the motion compensation prediction unit 306 acquires inter prediction information from the inter prediction mode determiner 305. A reference index and a motion vector are derived from the acquired inter prediction information, and the reference picture specified by the reference index in the decoded picture memory 104 is shifted from the same position as a picture signal of the block that is subjected to prediction processing by the amount of the motion vector. The picture signal of that position after the shift is acquired and thereafter a prediction signal is generated.
[0148] In a case where prediction is made from a single reference picture, such as when the inter prediction mode in the inter prediction is LO-prediction or L1-prediction, a prediction signal acquired from one reference picture is set as a motion compensation prediction signal. In a case where prediction mode is made from two reference pictures, such as when the inter prediction mode is BI prediction, a weighted averaging of prediction signals acquired from the two reference pictures is set as the motion compensation prediction signal. The acquired motion compensation prediction signal is supplied to the prediction method determiner 105. Here, the weighted averaging ratio in the bi-prediction is set to 1: 1. Alternatively, the weighted averaging may use another ratio. For example, the weighting ratio may be set such that the shorter the picture interval between the prediction target picture and the reference picture, the higher the weighting ratio. The calculation of the weighting ratio may also be performed using a correspondence table between the combination of the picture intervals and the weighting ratios.
[0149] The motion compensation prediction unit 406 has function similar to the motion compensation prediction unit 306 on the coding side. The motion compensation prediction unit 406 acquires inter prediction information from the normal motion vector predictor mode derivation unit 401, the normal merge mode derivation unit 402, the subblock motion vector predictor mode derivation unit 403, and the subblock merge mode derivation unit 404, via the switch 408. The motion compensation prediction unit 406 supplies the obtained motion compensation prediction signal to the decoded picture signal superimposer 207.
Inter prediction mode
[0150] The process of performing prediction from a single reference picture is defined as uni-prediction. Uni-prediction performs prediction of LO-prediction or L1-prediction using one of the two reference pictures registered in the reference lists L0 or L1.
[0151] Fig. 32 illustrates a case of uni-prediction in which the reference picture (RefL0Pic) of L0 is at a time before the target picture (CurPic). Fig. 33 illustrates a case of uni-prediction in which the reference picture of L0-prediction is at a time after the target picture. Similarly, uni-prediction can be performed by replacing the L0-prediction reference picture in Figs. 32 and 33 with an L1-prediction reference picture (RefL1Pic).
[0152] The process of performing prediction from two reference pictures is defined as bi-prediction. Bi-prediction performs prediction, expressed as BI prediction, using both LO-prediction and L1-prediction. Fig. 34 illustrates a case of bi-prediction in which an LO-prediction reference picture is at a time before the target picture and an L1-prediction reference picture is at a time after the target picture. Fig. 35 illustrates a case of bi-prediction in which the reference picture for LO-prediction and the reference picture for L1-prediction are at a time before the target picture. Fig. 36 illustrates a case of bi-prediction in which the reference picture for LO-prediction and the reference picture for L1-prediction are at a time after the target picture.
[0153] In this manner, it is possible to use prediction without limiting the relationship between the prediction type of L0/L1 and time such that L0 to the past direction and L1 to the future direction. Moreover, bi-prediction may perform each of LO-prediction and L1-prediction using a same reference picture. The determination whether to perform motion compensation prediction in the uni-prediction or the bi-prediction is made on the basis of information (for example, a flag) indicating whether to use the LO-prediction and whether to use the L1-prediction, for example.
Reference index
[0154] In the embodiment of the present invention, it is possible to select an optimal reference picture from a plurality of reference pictures in motion compensation prediction in order to improve motion compensation prediction accuracy. Therefore, the reference picture used in the motion compensation prediction is to be used as a reference index, and the reference index is coded in a bitstream together with the motion vector difference.
Motion compensation process based on normal motion vector predictor mode
[0155] As illustrated in the inter prediction unit 102 on the coding side in Fig. 16, in a case where inter prediction information by the normal motion vector predictor mode derivation unit 301 has been selected on the inter prediction mode determiner 305, the motion compensation prediction unit 306 acquires this inter prediction information from the inter prediction mode determiner 305, and derives an inter prediction mode, a reference index, and a motion vector of a target block and thereby generates a motion compensation prediction signal. The constructed motion compensation prediction signal is supplied to the prediction method determiner 105.
[0156] Similarly, as illustrated in the inter prediction unit 203 on the decoding side in Fig. 22, in a case where the switch 408 is connected to the normal motion vector predictor mode derivation unit 401 during the decoding process, the motion compensation prediction unit 406 acquires inter prediction information by the normal motion vector predictor mode derivation unit 401, and derives an inter prediction mode, a reference index, and a motion vector of current target block and thereby generates a motion compensation prediction signal. The constructed motion compensation prediction signal is supplied to the decoded picture signal superimposer 207.
Motion compensation processing based on normal merge mode
[0157] As illustrated in the inter prediction unit 102 on the coding side in Fig. 16, in a case where inter prediction information by the normal merge mode derivation unit 302 has been selected on the inter prediction mode determiner 305, the motion compensation prediction unit 306 acquires this inter prediction information from the inter prediction mode determiner 305, and derives an inter prediction mode, a reference index, and a motion vector of current target block, thereby generating a motion compensation prediction signal. The constructed motion compensation prediction signal is supplied to the prediction method determiner 105.
[0158] Similarly, as illustrated in the inter prediction unit 203 on the decoding side in Fig. 22, in a case where the switch 408 is connected to the normal merge mode derivation unit 402 during the decoding process, the motion compensation prediction unit 406 acquires inter prediction information by the normal merge mode derivation unit 402, and derives an inter prediction mode, a reference index, and a motion vector of current target block, thereby generating a motion compensation prediction signal. The constructed motion compensation prediction signal is supplied to the decoded picture signal superimposer 207.
Motion compensation process based on subblock motion vector predictor mode
[0159] As illustrated in the inter prediction unit 102 on the coding side in Fig. 16, in a case where inter prediction information by the subblock motion vector predictor mode derivation unit 303 has been selected on the inter prediction mode determiner 305, the motion compensation prediction unit 306 acquires this inter prediction information from the inter prediction mode determiner 305, and derives an inter prediction mode, a reference index, and a motion vector of current target block, thereby generating a motion compensation prediction signal. The constructed motion compensation prediction signal is supplied to the prediction method determiner 105.
[0160] Similarly, as illustrated in the inter prediction unit 203 on the decoding side in Fig. 22, in a case where the switch 408 is connected to the subblock motion vector predictor mode derivation unit 403 during the decoding process, the motion compensation prediction unit 406 acquires inter prediction information by the subblock motion vector predictor mode derivation unit 403, and derives an inter prediction mode, a reference index, and a motion vector of a target block, thereby generating a motion compensation prediction signal. The constructed motion compensation prediction signal is supplied to the decoded picture signal superimposer 207.
Motion compensation process based on subblock merge mode
[0161] As illustrated in the inter prediction unit 102 on the coding side in Fig. 16, in a case where inter prediction information by the subblock merge mode derivation unit 304 has been selected on the inter prediction mode determiner 305, the motion compensation prediction unit 306 acquires this inter prediction information from the inter prediction mode determiner 305, and derives an inter prediction mode, a reference index, and a motion vector of current target block, thereby generating a motion compensation prediction signal. The constructed motion compensation prediction signal is supplied to the prediction method determiner 105.
[0162] Similarly, as illustrated in the inter prediction unit 203 on the decoding side in Fig. 22, in a case where the switch 408 is connected to the subblock merge mode derivation unit 404 during the decoding process, the motion compensation prediction unit 406 acquires inter prediction information by the subblock merge mode derivation unit 404, and derives an inter prediction mode, a reference index, and a motion vector of current target block, thereby generating a motion compensation prediction signal. The constructed motion compensation prediction signal is supplied to the decoded picture signal superimposer 207.
Motion compensation process based on affine transform prediction
[0163] In the normal motion vector predictor mode and the normal merge mode, motion compensation using an affine model is usable based on the following flags. The following flags are reflected in the following flags on the basis of inter prediction conditions determined by the inter prediction mode determiner 305 in the coding process, and are coded in the bitstream. In the decoding process, whether to perform motion compensation using the affine model on the basis of the following flags in the bitstream is specified.
[0164] sps_affine_enabled_flag indicates whether motion compensation using an affine model is usable in inter prediction. When sps_affine_enabled_flag is 0, the process is suppressed so as not to perform motion compensation by the affine model in units of sequence. Moreover, inter_affine_flag and cu_affine_type_flag are not transmitted in the coding block (CU) syntax of a coding video sequence. When sps_affine_enabled_flag is 1, motion compensation by an affine model is usable in the coding video sequence.
[0165] sps_affine_type_flag indicates whether motion compensation using a 6-parameter affine model is usable in inter prediction. When sps_affine_type_flag is 0, the process is suppressed so as not to perform motion compensation using a 6-parameter affine model. Moreover, cu_affine_type_flag is not transmitted in the CU syntax of the coding video sequence. When sps_affine_type_flag is 1, motion compensation based on a 6-parameter affine model is usable in a coding video sequence. In a case where sps_affine_type_flag does not exist, it shall be 0.
[0166] In a case of decoding a P or B slice, when inter_affine_flag is 1 in the current target CU, a motion compensation using an affine model is used in order to generate a motion compensation prediction signal of the current target CU. When inter_affine_flag is 0, the affine model is not used for the current target CU. In a case where inter_affine_flag does not exist, it shall be 0.
[0167] In a case of decoding a P or B slice, when cu_affine_type_flag is 1 in the current CU, a motion compensation using a 6-parameter affine model is used in order to generate a motion compensation prediction signal of the current CU. When cu_affine_type_flag is 0, motion compensation using a four-parameter affine model is used to generate a motion compensation prediction signal of the CU currently being processed.
[0168] A reference index and a motion vector are derived in units of subblocks in the motion compensation based on the affine model. Accordingly, a motion compensation prediction signal is generated using the reference index and the motion vector to be processed in subblock units.
[0169] The four-parameter affine model is a mode in which a motion vector of a subblock is derived from four parameters of a horizontal component and a vertical component of each of motion vectors of two control points, and motion compensation is performed in units of subblocks.
Triangle merge mode
[0170] The triangle merge mode is a type of merge mode, in which the coding/decoding block is split into diagonal partitions to perform motion compensation prediction. The triangle merge mode is a type of geometric division merge mode in which the coding/decoding block is split into blocks having a non-rectangular shape. In the geometric division merge mode, this corresponds to a mode in which the coding/decoding block is split into two right triangles by a diagonal line.
[0171] The geometric division merge mode is expressed by a combination of two parameters, for example, an index (angleIdx) indicating a division angle and an index (distanceIdx) indicating a distance from the center of the coding block. As an example, 64 patterns are defined as the geometric division merge mode, and fixed-length encoding is performed. Of the 64 patterns, two modes, in which the index indicating a division angle indicates an angle forming a diagonal line of the coding block (for example, 45 degrees (angleIdx = 4 in a configuration in which 360 degrees are represented by 32 divisions) or 135 degrees (angleIdx = 12 in a configuration in which 360 degrees are represented by 32 divisions)) and the index indicating a distance from the center of the coding block is minimum (distanceIdx = 0, indicating that the division boundary passes through the center of the coding block), indicate that the coding block is split by a diagonal line, and correspond to the triangle merge mode.
[0172] The triangle merge mode will be described with reference to Figs. 38A and 38B. Figs. 38A and 38B illustrate an example of prediction of a 16 × 16 coding/decoding blocks of the triangle merge mode. The coding/decoding block of the triangle merge mode is split into 4 × 4 subblocks, and each of subblock is assigned to three partitions, namely, uni-prediction partition 0 (UNIO), uni-prediction partition 1 (UNI1), and bi-prediction partition 2 (BI). Here, subblocks above a diagonal line are assigned to partition 0, subblocks below the diagonal line are assigned to partition 1, and subblocks on the diagonal line are assigned to partition 2. When merge_triangle_split_dir is 0, partitions are assigned as illustrated in Fig. 38A, and when merge_triangle_split_dir is 1, partitions are assigned as illustrated in Fig. 38B.
[0173] Uni-prediction motion information designated by merge triangle index 0 is used for motion compensation prediction of partition 0. Uni-prediction motion information designated by merge triangle index 1 is used for motion compensation prediction of partition 1. Bi-prediction motion information combining uni-prediction motion information designated by merge triangle index 0 and uni-prediction motion information designated by merge triangle index 1 is used for motion compensation prediction of partition 2.
[0174] Here, the uni-prediction motion information is a set of a motion vector and a reference index, while the bi-prediction motion information is formed with two sets of a motion vector and a reference index. The motion information represents either uni-prediction motion information or bi-prediction motion information.
[0175] The merging candidate selectors 347 and 447 use the derived merging candidate list mergeCandList as a triangle merging candidate list triangleMergeCandList.
[0177] First, a merging candidate list mergeCandList is used as a triangle merging candidate list triangleMergeCandList (step S3501).
[0178] Subsequently, a merging candidate having a motion information list L0 is prioritized, and uni-prediction motion information of a merge triangle partition 0 is derived (step S3502) .
[0179] Subsequently, a merging candidate having a motion information list L1 is prioritized, and uni-prediction motion information of a merge triangle partition 1 is derived (step S3503).
[0180] Note that step S3502 and step S3503 can be derived in random order and can also be processed in parallel.
[0181] Fig. 41 is a flowchart illustrating derivation of uni-prediction motion information of a merge triangle partition 0 according to the present embodiment.
[0182] First, for an M-th candidate in the derived merging candidate list mergeCandList, determination is made as to whether a candidate M has motion information of the motion information list L0 (step S3601). In a case where the candidate M has the motion information of the motion information list L0, the motion information of the motion information list L0 of the candidate M is set as a triangle merging candidate (step S3602). For candidates M (M = 0, 1,..., numMergeCand-1), steps S3601 and step S3602 are performed in ascending order, and triangle merging candidates are additionally derived.
[0183] Subsequently, for the M-th candidate in the derived merging candidate list mergeCandList, determination is made as to whether the candidate M has motion information of the motion information list L1 (step S3603). In a case where the candidate M has the motion information of the motion information list L1, the motion information of the motion information list L1 of the candidate M is set as a triangle merging candidate (step S3604). For candidates M (M = numMergeCand-1,..., 1, 0), steps S3603 and step S3604 are performed in descending order, and triangle merging candidates are additionally derived.
[0184] Fig. 42 is a flowchart illustrating derivation of uni-prediction motion information of a merge triangle partition 1 according to the present embodiment.
[0185] First, for an M-th candidate in the derived merging candidate list mergeCandList, determination is made as to whether a candidate M has motion information of the motion information list L1 (step S3701). In a case where the candidate M has the motion information of the motion information list L1, the motion information of the motion information list L1 of the candidate M is set as a triangle merging candidate (step S3702). For candidates M (M = 0, 1,..., numMergeCand-1), steps S3701 and step S3702 are performed in ascending order, and triangle merging candidates are additionally derived.
[0186] Subsequently, for the M-th candidate in the derived merging candidate list mergeCandList, determination is made as to whether the candidate M has motion information of the motion information list L0 (step S3703). In a case where the candidate M has the motion information of the motion information list L0, the motion information of the motion information list L0 of the candidate M is set as a triangle merging candidate (step S3704). For candidates M (M =numMergeCand-1,..., 1, 0), steps S3703 and step S3704 are performed in descending order, and triangle merging candidates are additionally derived.
[0187] The merging candidate selector 347 on the coding side acquires motion information from the derived triangle merging candidate list triangleMergeCandList and calculates code amounts and distortion amounts.
[0188] The merging candidate selector 347 compares calculated pluralities of code amounts and distortion amounts, thereby selecting coding block splitting directions and triangle merging candidates of divided partitions. In a case where encoding is performed using the triangle merge mode, the merging candidate selector 347 supplies selected information (the coding block splitting directions merge_triangle_split_dir and merge triangle indexes merge_triangle_idx0, merge_triangle_idx1 indicating the triangle merging candidates of the divided partitions) and inter prediction information of the triangle merging candidates to the motion compensation prediction unit 306. The bit strings coding unit 108 encodes the selected information.
[0189] On the other hand, in the case of the triangle merge mode, the merging candidate selector 447 on the decoding side selects triangle merging candidates based on decoded information (the coding block splitting directions merge_triangle_split_dir and the merge triangle indexes merge_triangle_idx0 and merge_triangle_idx1 indicating the triangle merging candidates of the divided partitions) and supplies inter prediction information of the selected triangle merging candidates to the motion compensation prediction unit 406.
[0190] In the case of the triangle merge mode, the motion compensation prediction units 306 and 406 perform weighted averaging described below. In the case of luminance, the motion compensation prediction units 306 and 406 calculate nCbR = (nCbW > nCbH) ? (nCbW / nCbH) : (nCbH / nCbW) with respect to width nCbW and height nCbH of the coding block. Then, in a position (x, y) in the coding block, weight wValue in the case of Fig. 38A is calculated as wValue = (nCbW > nCbH) ?
On the other hand, the weight wValue in the case of Fig. 38B is calculated as wValue = (nCbW > nCbH) ?
Further, the motion compensation prediction units 306 and 406 calculate shift1 = max(5, 17 - bitDepth)
[0191] offset1 = 1 << (shift1 - 1) with respect to a bit number bitDepth. Then, a result pbSamples of the weighted averaging is calculated as pbSamples = Clip3(0, (1 << bitDepth) - 1,
Here, predSamplesLA is a pixel value motion-compensated using a motion vector mvLA and predSamplesLB is a pixel value motion-compensated using mvLB.
Storage process in the coding information storage memory
[0192] The inter prediction information obtained in the triangle merge mode is stored in the coding information storage memory so that the inter prediction information can be referred to as inter prediction information neighboring a target block when coding and decoding are performed. A storage process in the coding information storage memory is performed in 4x4 subblock units and the inter prediction information specified by the partitions is stored.
[0193] The specified inter prediction information is inter prediction information of a partition 0(UNI0) of unit-prediction and a partition 1(UNI1) of uni-prediction. A partition 2(BI) of bi-prediction is obtained by using the inter prediction information of UNIO and UNI1.
[0194] For subblocks on diagonal lines split as the partitions, in a case where weighting illustrated in Fig. 43 is performed, a partition 2 illustrated in Fig. 44 is conceivable. Fig. 44A illustrates a case where a region where the weighting illustrated in Fig. 43A is performed is set as the partition 2. Similarly, Fig. 44B corresponds to Fig. 43B. Fig. 44C illustrates a case where subblocks for which the weighting is performed in Fig. 43A is set as the partition 2. Similarly, Fig. 44D corresponds to Fig. 43B. Fig. 44E illustrates a case where subblocks for the entirety of which the weighting is performed in Fig. 43A is set as the partition 2. Similarly, Fig. 44F corresponds to Fig. 43B.
[0195] A region of the partition 2 illustrated in Fig. 44 can also be stored as BI as illustrated in Fig. 45. However, in this embodiment, for subblocks for a part of which weighting is performed as illustrated in Fig. 46, inter prediction information belonging to a partition having a larger weighting value is stored as uni-prediction (UNIY). Here, Y is 0 or 1.
[0196] Further, as illustrated in Fig. 47 and Fig. 48, subblocks originally stored as BI are stored as uni-prediction using inter prediction information of a predetermined partition. Fig. 47A and Fig. 47B are a case where UNIO is the predetermined partition. Fig. 48A and Fig. 48B are a case where UNI1 is the predetermined partition.
[0197] As described in this embodiment, by storing a partition 2(BI) of bi-prediction in the coding information storage memory as UNIY of uni-prediction, it is possible to reduce a memory amount for storage. Further, since motion information of the motion information list L0 and the motion information list L1 is stored in the coding information storage memory without being converted into BI prediction, it is possible to reduce a processing amount. Further, it is possible to reduce a processing amount in a case where the inter prediction information specified by the triangle merge mode is referred to and used in coding and decoding thereafter.
[0198] By selecting UNI1 as the predetermined partition as illustrated in Fig. 48A and Fig. 48B, in a case where a block in the neighbor of the right of a triangle merge mode illustrated in Fig. 48B is in a triangle merge mode illustrated in Fig. 48A, continuity of motion information to be stored can also be maintained in a region where the two triangle merge modes continue. Therefore, coding efficiency is improved compared with selecting UNIO as the predetermined partition as illustrated in Fig. 47A and Fig. 47B.
[0199] Selecting UNI1 as the predetermined partition as illustrated in Fig. 48A and Fig. 48B can also maintain continuity of processing because motion information of partitions is accumulated after coding/decoding. Therefore, a processing amount is reduced compared with selecting UNIO as the predetermined partition as illustrated in Fig. 47A and Fig. 47B.
[0200] By fixing, as L0 prediction, uni-prediction to be saved as UNI1, since the L0 prediction can be treated in the same manner as P slice, a processing amount is reduced compared with a case where uni-prediction saved as L1 prediction.
[0201] In all the embodiments described above, a plurality of technologies may be combined with each other.
[0202] In all the embodiments described above, the bitstream output from the picture coding device has a specific data format so as to be decoded following the coding method used in the embodiment. Furthermore, the picture decoding device corresponding to the picture coding device is capable of decoding the bitstream of the specific data format.
[0203] In a case where a wired or wireless network is used to exchange a bitstream between the picture coding device and the picture decoding device, the bitstream may be converted to a data format suitable for the transmission form of the communication channel in transmission. In this case, there are provided a transmission device that converts the bitstream output from the picture coding device into coded data in a data format suitable for the transmission form of the communication channel and transmits the coded data to the network, and a reception device that receives the coded data from the network to be restored to the bitstream and supplies the bitstream to the picture decoding device. The transmission device includes memory that buffers a bitstream output from the picture coding device, a packet processing unit that packetizes the bitstream, and a transmitter that transmits packetized coded data via a network. The reception device includes a receiver that receives a packetized coded data via a network, memory that buffers the received coded data, and a packet processing unit that packetizes coded data to construct a bitstream and supplies the constructed bitstream to the picture decoding device.
[0204] Moreover, a display unit that displays a picture decoded by the picture decoding device may be added, as a display device, to the configuration. In that case, the display unit reads out a decoded picture signal constructed by the decoded picture signal superimposer 207 and stored in the decoded picture memory 208, and displays the signal on the screen.
[0205] Moreover, an imaging unit may be added to the configuration so as to function as an imaging device by inputting a captured picture to the picture coding device. In that case, the imaging unit inputs the captured picture signal to the block split unit 101.
[0206] Fig. 37 illustrates an example of a hardware configuration of the coding-decoding device according to the present embodiment. The coding-decoding device includes the configurations of the picture coding device and the picture decoding device according to the embodiments of the present invention. A coding-decoding device 9000 includes a CPU 9001, a codec IC 9002, an I/O interface 9003, memory 9004, an optical disk drive 9005, a network interface 9006, and a video interface 9009, in which individual units are connected by a bus 9010.
[0207] A picture encoder 9007 and a picture decoder 9008 are typically implemented as a codec IC 9002. The picture coding process of the picture coding device according to the embodiments of the present invention is executed by the picture encoder 9007. The picture decoding process in the picture decoding device according to the embodiment of the present invention is executed by the picture decoder 9008. The I/O interface 9003 is implemented by a USB interface, for example, and connects to an external keyboard 9104, mouse 9105, or the like. The CPU 9001 controls the coding-decoding device 9000 on the basis of user's operation input via the I/O interface 9003 so as to execute operation desired by the user. The user's operations on the keyboard 9104, the mouse 9105, or the like include selection of which function of coding or decoding is to be executed, coding quality setting, input/output destination of a bitstream, input/output destination of a picture, or the like.
[0208] In a case where the user desires operation of reproducing a picture recorded on a disk recording medium 9100, the optical disk drive 9005 reads out a bitstream from the inserted disk recording medium 9100, and transmits the readout bitstream to the picture decoder 9008 of the codec IC 9002 via the bus 9010. The picture decoder 9008 executes a picture decoding process in the picture decoding device according to the embodiments of the present invention on the input bitstream, and transmits the decoded picture to the external monitor 9103 via the video interface 9009. The coding-decoding device 9000 has a network interface 9006, and can be connected to an external distribution server 9106 and a mobile terminal 9107 via a network 9101. In a case where the user desires to reproduce a picture recorded on the distribution server 9106 or the mobile terminal 9107 instead of the picture recorded on the disk recording medium 9100, the network interface 9006 obtains a bitstream from the network 9101 instead of reading out a bitstream from the input disk recording medium 9100. In a case where the user desires to reproduce the picture recorded in the memory 9004, the picture decoding processing is performed by the picture decoding device according to the embodiments of the present invention on the bitstream recorded in the memory 9004.
[0209] In a case where the user desires to perform operation of coding a picture captured by an external camera 9102 and recording the picture in the memory 9004, the video interface 9009 inputs the picture from the camera 9102, and transmits the picture to the picture encoder 9007 of the codec IC 9002 via the bus 9010. The picture encoder 9007 executes the picture coding process by the picture coding device according to the embodiment of the present invention on a picture input via the video interface 9009 and thereby creates a bitstream. Subsequently, the bitstream is transmitted to the memory 9004 via the bus 9010. In a case where the user desires to record a bitstream on the disk recording medium 9100 instead of the memory 9004, the optical disk drive 9005 writes the bitstream on the inserted disk recording medium 9100.
[0210] It is also possible to implement a hardware configuration having a picture coding device and not having a picture decoding device, or a hardware configuration having a picture decoding device and not having a picture coding device. Such a hardware configuration is implemented by replacing the codec IC 9002 with the picture encoder 9007 or the picture decoder 9008.
[0211] The above-described process related to coding and decoding may naturally be implemented as a transmission, storage, and reception device using hardware, and alternatively, the process may be implemented by firmware stored in read only memory (ROM), flash memory, or the like, or by software provided for a computer or the like. The firmware program and the software program may be provided by being recorded on a recording medium readable by a computer or the like, may be provided from a server through a wired or wireless network, or may be provided through data broadcasting by terrestrial or satellite digital broadcasting.
[0212] The present invention has been described with reference to the present embodiments. The above-described embodiment has been described merely for exemplary purposes. Rather, it can be readily conceived by those skilled in the art that various modification examples may be made by making various combinations of the above-described components or processes, which are also encompassed in the technical scope of the present invention.
[0213] The present invention can be used for picture coding and decoding techniques that split a picture into blocks to perform prediction.
- 100
- picture coding device
- 101
- block split unit
- 102
- inter prediction unit
- 103
- intra prediction unit
- 104
- decoded picture memory
- 105
- prediction method determiner
- 106
- residual generation unit
- 107
- orthogonal transformer/quantizer
- 108
- bit strings coding unit
- 109
- inverse quantizer/inverse orthogonal transformer
- 110
- decoded picture signal superimposer
- 111
- coding information storage memory
- 200
- picture decoding device
- 201
- bit strings decoding unit
- 202
- block split unit
- 203
- inter prediction unit
- 204
- intra prediction unit
- 205
- coding information storage memory
- 206
- inverse quantizer/inverse orthogonal transformer
- 207
- decoded picture signal superimposer
- 208
- decoded picture memory
1. A moving picture coding device using a triangle merge mode, the moving picture coding
device comprising:
a triangle merging candidate list constructor structured to construct a triangle merging candidate list including spatial merging candidates;
a first triangle merging candidate selector structured to select, from the triangle merging candidate list, a first triangle merging candidate that is used for uni-prediction partition 0 which consists of subblocks above a diagonal line of triangle partition splitting a coding block; and
a second triangle merging candidate selector structured to select, from the triangle merging candidate list, a second triangle merging candidate that is used for uni-prediction partition 1 which consists of subblocks below the diagonal line, wherein
in bi-prediction partition 2 which consists of subblocks on the diagonal line where motion compensation by weighted averaging by the first triangle merging candidate and the second triangle merging candidate is performed, uni-prediction motion information of the second triangle merging candidate is stored in a coding information storage memory so that the uni-prediction motion information can be referred to as inter prediction information for a coding block neighboring the coding block without storing bi-prediction information which is converted from motion information of list L0 and motion information of list L1.
2. A moving picture coding method using a triangle merge mode, the moving picture coding
method comprising:
a triangle merging candidate list constructing step for constructing a triangle merging candidate list including spatial merging candidates;
a first triangle merging candidate selecting step for selecting, from the triangle merging candidate list, a first triangle merging candidate that is used for uni-prediction partition 0 which consists of subblocks above a diagonal line of triangle partition splitting a coding block; and
a second triangle merging candidate selecting step for selecting, from the triangle merging candidate list, a second triangle merging candidate that is used for uni-prediction partition 1 which consists of subblocks below the diagonal line, wherein
in bi-prediction partition 2 which consists of subblocks on the diagonal line where motion compensation by weighted averaging by the first triangle merging candidate and the second triangle merging candidate is performed, uni-prediction motion information of the second triangle merging candidate is stored in a coding information storage memory so that the uni-prediction motion information can be referred to as inter prediction information for a coding block neighboring the coding block without storing bi-prediction information which is converted from motion information of list L0 and motion information of list L1.
3. A moving picture coding program using a triangle merge mode, the moving picture coding
program causing a computer to execute:
a triangle merging candidate list constructing step for constructing a triangle merging candidate list including spatial merging candidates;
a first triangle merging candidate selecting step for selecting, from the triangle merging candidate list, a first triangle merging candidate that is used for uni-prediction partition 0 which consists of subblocks above a diagonal line of triangle partition splitting a coding block; and
a second triangle merging candidate selecting step for selecting, from the triangle merging candidate list, a second triangle merging candidate that is used for uni-prediction partition 1 which consists of subblocks below the diagonal line, wherein
in bi-prediction partition 2 which consists of subblocks on the diagonal line where motion compensation by weighted averaging by the first triangle merging candidate and the second triangle merging candidate is performed, uni-prediction motion information of the second triangle merging candidate is stored in a coding information storage memory so that the uni-prediction motion information can be referred to as inter prediction information for a coding block neighboring the coding block without storing bi-prediction information which is converted from motion information of list L0 and motion information of list L1.
4. A moving picture decoding device using a triangle merge mode, the moving picture decoding
device comprising:
a triangle merging candidate list constructor structured to construct a triangle merging candidate list including spatial merging candidates;
a first triangle merging candidate selector structured to select, from the triangle merging candidate list, a first triangle merging candidate that is used for uni-prediction partition 0 which consists of subblocks above a diagonal line of triangle partition splitting a decoding block; and
a second triangle merging candidate selector structured to select, from the triangle merging candidate list, a second triangle merging candidate that is used for uni-prediction partition 1 which consists of subblocks below the diagonal line, wherein
in bi-prediction partition 2 which consists of subblocks on the diagonal line where motion compensation by weighted averaging by the first triangle merging candidate and the second triangle merging candidate is performed, uni-prediction motion information of the second triangle merging candidate is stored in a decoding information storage memory so that the uni-prediction motion information can be referred to as inter prediction information for a decoding block neighboring the decoding block without storing bi-prediction information which is converted from motion information of list L0 and motion information of list L1.
5. A moving picture decoding method using a triangle merge mode, the moving picture decoding
method comprising:
a triangle merging candidate list constructing step for constructing a triangle merging candidate list including spatial merging candidates;
a first triangle merging candidate selecting step for selecting, from the triangle merging candidate list, a first triangle merging candidate that is used for uni-prediction partition 0 which consists of subblocks above a diagonal line of triangle partition splitting a decoding block; and
a second triangle merging candidate selecting step for selecting, from the triangle merging candidate list, a second triangle merging candidate that is used for uni-prediction partition 1 which consists of subblocks below the diagonal line, wherein
in bi-prediction partition 2 which consists of subblocks on the diagonal line where motion compensation by weighted averaging by the first triangle merging candidate and the second triangle merging candidate is performed, uni-prediction motion information of the second triangle merging candidate is stored in a decoding information storage memory so that the uni-prediction motion information can be referred to as inter prediction information for a decoding block neighboring the decoding block without storing bi-prediction information which is converted from motion information of list L0 and motion information of list L1.
6. A moving picture decoding program using a triangle merge mode, the moving picture
decoding program causing a computer to execute:
a triangle merging candidate list constructing step for constructing a triangle merging candidate list including spatial merging candidates;
a first triangle merging candidate selecting step for selecting, from the triangle merging candidate list, a first triangle merging candidate that is used for uni-prediction partition 0 which consists of subblocks above a diagonal line of triangle partition splitting a decoding block; and
a second triangle merging candidate selecting step for selecting, from the triangle merging candidate list, a second triangle merging candidate that is used for uni-prediction partition 1 which consists of subblocks below the diagonal line, wherein
in bi-prediction partition 2 which consists of subblocks on the diagonal line where motion compensation by weighted averaging by the first triangle merging candidate and the second triangle merging candidate is performed, uni-prediction motion information of the second triangle merging candidate is stored in a decoding information storage memory so that the uni-prediction motion information can be referred to as inter prediction information for a decoding block neighboring the decoding block without storing bi-prediction information which is converted from motion information of list L0 and motion information of list L1.
REFERENCES CITED IN THE DESCRIPTION
This list of references cited by the applicant is for the reader's convenience only. It does not form part of the European patent document. Even though great care has been taken in compiling the references, errors or omissions cannot be excluded and the EPO disclaims all liability in this regard.
Patent documents cited in the description